Deontic Logic: Introductory and Systematic Readings

DEONTIC LOGIC: INTRODUCTORY AND SYSTEMATIC READINGS
A PALLAS PAPERBACK I 20
DEONTIC LOGIC:
INTRODUCTORY AND
SYSTEMATIC READINGS
Edited by
RISTO HILPINEN
Department of Philosophy, University of Turku, Finlllnd
~p~
~ paperbaCkS
D. REIDEL PUBLISHING COMPANY

DORDRECHT : HOLLAND / BOSTON: U.S.A.
LONDON: ENGLAND
Library of Congress Cataloging in Publication Data
Main entry under title:
Deontic logic.
(Synthese library; v. 33)

Includes bibliographies and indexes.
Contents: Deontic logic / Dagimn F011esdal and Risto Hilpinen -
New foundations for ethical theory / Stig Kanger - Some main problems
of deontic logic / laakko Hintikka - [etc.]
1. Deontic logic - Adresses, essays, lectures. I. Hilpinen, Risto.
BC145.D45 1981 160 72-135103
AACR2
ISBN-13: 978-90-277-1302-5 e-ISBN-13: 978-94-010-3146-2
DOl: 10.1 007/978-94-010-3146-2
First published in 1971

Reprinted in 1981, with new Introduction
Published by D. Reidel Publishing Company
P.O. Box 17,3300 AA Dordrecht, Holland
Sold and distributed in the U.S.A. and Canada

by Kluwer Boston Inc.,
190 Old Derby Street, Hingham, MA 02043, U.S.A.
In all other countries, sold and distributed

by Kluwer Academic Publishers Group
P.O. Box 322, 3300 AH Dordrecht, Holland
D. Reidel Publishing Company is a member of the Kluwer Group
First published in 1971 in hardbound edition

by Reidel in the series Synthese Library. Volume 33
All Rights Reserved

Copyright © 1971, 1981 by D. Reidel Publishing Company, Dordrecht, Holland
and copyright owners as specified on appropriate pages within

No part of the material protected by this copyright notice may be reproduced or
utilized in any form or by any means, electronic or mechanical,
including photocopying, recording or by any information storage and
retrieval system, without written permission from the copyright owner
TABLE OF CONTENTS
PREFACE TO THE SECOND IMPRESSION VII
PREFACE TO THE FIRST EDITION IX
INTRODUCTION TO THE SECOND IMPRESSION xi
DAGFINN FI/>LLESDAL and RISTO HILPINEN / Deontic Logic:

An Introduction
STIG KANGER / New Foundations for Ethical Theory 36
JAAKKO HINTIKKA/Some Main Problems of Deontic Logic 59
GEORG HENRIK VON WRIGHT / A New System of Deontic Logic 105
BENGT HANSSON / An Analysis of Some Deontic Logics 121
KRISTER SEGER BERG / Some Logics of Commitment and

Obligation 148
GEORG HENRIK VON WRIGHT / Deontic Logic and the Theory of

Conditions 159
INDEX OF NAMES 178
INDEX OF SUBJECTS 180

PREFACE TO
THE SECOND IMPRESSION
This second impression of Deontic Logic: Introductory and Systematic

Readings is published simultaneously with a companion volume entitled
New Studies in Deontic Logic (D. Reidel Publishing Company, Dordrecht
1981). The papers published in the present collection outline the 'basic' or
'standard' system of deontic logic, discuss its applications to moral philo-
sophy, and analyse the concepts of commitment and conditional ob-
ligation. New Studies contains ten previously unpublished papers on the
philosophical foundations and various special topics of the logic of norms:
the foundations of deontic logic and action theory, problems of normative
conflict, the interrelations between normative and temporal notions, and
the history of deontic logic. The two volumes together form a compre-
hensive survey of the main problems and results of contemporary deontic
logic.
The present impression contains a new Introduction which discusses the
relationship of the papers published here to some new developments in the
field; otherwise the book is reprinted (apart from the correction of some
misprints) in the original form.
PREFACE TO
THE FIRST EDITION
In its modern form deontic logic, or the logic of normative concepts, has
been studied for about twenty years. The relevant literature consists
mainly of papers in various journals and publications, many of which
are not easily accessible to students of philosophy. Introductory text-
books have so far been missing. This volume has a two-fold purpose:
first, it brings together some important contributions to deontic logic,
and secondly, the papers published here have been selected in such a way
that they jointly can serve as an introduction to the problems and methods
of deontic logic. The essay~ by Fl1Illesdal and Hilpinen, Hintikka (partly),
and Segerberg were written specifically for this book; the others have been
published before. Some of these papers appear here in a revised form.
Most papers require some familiarity with elementary logic.
In the study of deontic logic, as well as other branches of what is now
commonly called 'philosophical logic', we can distinguish between two
main aspects. On one hand, deontic logicians have constructed formal
theories of various normative concepts, and on the other, they have tested
the adequacy of these theories by applying them to analysis of ethical
discussion. These applications may also involve criticism of ethical or
meta-ethical views. In this book, the former aspect is emphasized in
Kanger's paper, and the latter, e.g., in Hintikka's essay. A characteristic
feature of this book is the prevalence of semantical methods. These
methods are most likely to lead to significant further developments, not
only in deontic logic, but in philosophical logic and analysis in general.
All the papers included in this volume are by Scandinavian authors.
This book reflects the strength of interest in deontic logic and related
subjects in Scandinavia. In fact, Scandinavian philosophers have done a
great deal of the pioneering work in this field. The modern development
of deontic logic was initiated in the early 1950's by G. H. von Wright,
whose work stimulated most of the subsequent discussion of the subject.
Semantical theories of deontic notions were also presented first by
Scandinavian philosophers, Jaakko Hintikka and Stig Kanger (1957).
x PREFACE
I wish to express my thanks to all contributors to this volume for their

collaboration, and especially to Professor laakko Hintikka, Editor of
Synthese Library, for his help in editing this book.
THE EDITOR
INTRODUCTION TO
THE SECOND IMPRESSION
The articles included in this collection represent what may be called the
standard modal approach to deontic logic (the logic of normative concepts),
in which deontic logic is treated as a branch of modal logic, and the
normative concepts of obligation, permission (permissibility) and prohibi-
tion are regarded as analogous to the 'alethic' modalities necessity, possi-
bility and impossibility. In his recent paper [16] Simo Knuuttila has
shown that this approach can be traced back to late medieval philosophy.
Several 14th century philosophers observed the analogies between deontic
and alethic modalities and 4iscussed the deontic interpretations of various
laws of modal logic. A relatively simple deontic system of this kind (called
the system D or K D; cf. Lemmon and Scott [17], pp. 50-51, Chellas [10], p.
131) is obtained by adding to propositional logic two deontic axioms (or
axiom schemata),
(K) O(A :::J B) :::J (OA :::J OB)
and
(D) OA :::J ~ 0 ~ A,
where '0' is the obligation operator, and the deontic variant of the 'rule of
necessitation'
(0) From A, to infer ~A.
This system is closely related to the familiar alethic system T: the latter is
obtained from D by replacing '0' by its alethic counterpart '0' and
by strengthening the schema (D) into
(T) 0 A :::J A.
The system D is often called 'the standard system of deontic logic'. (The
standard system can be formulated in different ways; cf. pp. 13 and 127-128
of this volume.) More generally, any deontic system which includes the
system D may be termed a 'standard system'.
xii INTRODUCTION TO THE SECOND IMPRESSION
In the present volume, the papers by Stig Kanger and laakko Hintikka
outline the basic semantics of deontic statements, and laakko Hintikka
also shows how deontic logic can be brought to bear on various problems
and issues of moral philosophy, e.g. the is-ought question, the 'sollen-
konnen' principle, and the distinction between absolute obligations and
prima facie obligations. Georg Henrik von Wright (in 'A New System of
Deontic Logic'), Bengt Hansson, and Krister Segerberg discuss the pro-
blems of conditional obligation and commitment: in their contributions they
develop systems of dyadic deontic logic, in which the dependence of the
normative status of an act on the circumstances in which it is performed or
on the agent's earlier acts is expressed in terms of a dyadic obligation
operator (O-operator). In his second paper ('Deontit Logic and the Theory
of Conditions'), Georg Henrik von Wright suggests that deontic logic can
be construed as a fragment of the modal logic of (necessary and sufficient)
conditions, and argues that this way of looking at deontic logic leads to
illuminating analyses of the concepts of commitment and strong (disjunc-
tive) permission.
Much of the recent work on the logic of norms has centered on certain
'paradoxes', that is, arguments and examples which seem to conflict with
the basic principles of the standard approach. For example, Georg Henrik
von Wright's.and Bengt Hansson's work on conditional obligation -(re-
ported in the present volume) has been motivated by the 'paradoxes of
commitment' (see F~llesdal and Hilpinen's paper in this volume, pp. 23-24)
and by Roderick M. Chisholm's 'paradox of contrary-to-duty obligation'
(see pp. 24-25, 105 and 132-133). Chisholm's example is one of a group
involving conflicting obligations. The schema (D) or the principle of consis-
tency of the standard system excludes the possibility of genuine normative
conflicts; thus examples involving seemingly conflicting obligations pre-
sent a problem for the standard approach. In the case of Chisholm's
example, von Wright and Hansson solve the problem by relativising the
concept of obligation to circumstances: they assume that mutually incom-
patible obligations are relative to different conditions or circumstances. In
the 1970's the logic of conditional obligation has been studied by Bas C.
van Fraassen [25], David Lewis ([ 18], pp. 96-104, [19]), and by Azizah al-
Hibri, whose monograph [3] contains a survey of various deontic para-
doxes and a critical review of the theories of conditional obligation pro-
posed by von Wright, Hansson, van Fraassen, Lewis, and others. Brian F.
INTRODUCTION TO THE SECOND IMPRESSION X111
Chellas ([9] and [to], pp. 275-276) and Peter Mott [21] have presented
especially perspicuous analyses of conditional obligation in which the
ideas of obligation and conditionality are separated from each other, and
conditional obligations are defined in terms of the standard (monadic)
obligation operator and a dyadic conditional operator. Such an analysis
resembles that presented by Krister Seger berg in the present volume,
except that the conditional connective employed by Chellas and Mott is
not a strict or necessary conditional, but a variably strict conditional (in the
sense of David Lewis [18]; see p. 13), and consequently the concept of
conditional obligation defined by Chellas and Mott does not satisfy the
(objectionable) augmentation principle
(A com C) :::> «A & B) com C)
(where 'com' stands for the concept of commitment) entailed by

Segerberg's analysis (see p. 156 below).
Some philosophers have tried to reconcile apparent moral conflicts with
the consistency principle by relativising the concept of obligation to time: it
has been suggested that mutually incompatible obligations can only con-
cern an agent at different times. For example, Chisholm's 'contrary-to-
duty obligations' are actualized only after the agent has failed to comply
with his primary duties - or after his failure to fulfil his primary obligations
has become unavoidable (see Patricia Greenspan [11], pp. 265-267). Moral
conflicts aye not the only reason for studying the interrelations of deontic
and temporal modalities. If deontic sentences are used for the purpose of
directing and regulating people's conduct, they are essentially 'forward-
looking': deontic sentences about the past can be used only for judging the
agent and his actions. Richmond H. Thomason ([23] and [24]) has distin-
guished between the deliberative and the judgmental use of deontic sen-
tences, and discussed the relevance of the temporal relativity of obligations
(and ought-sentences) to moral deliberation. Systems of deontic tense logic
have also been developed by Brian F. Chellas ([8]; [10], pp. 198-200) and
by Lennart Aqvist and Jaap Hoepelman [4] (among others).
It is not obvious that all moral conflicts can be 'explained away' in the
ways described above: perhaps some of them are genuine and not merely
apparent (or prima facie) conflicts. To countenance this possibility, some
philosophers have developed deontic systems in which the consistency
XIV INTRODUCTION TO THE SECOND IMPRESSION
principle (D) does not hold. The axiom schema (K) implies (by rule (0) and
propositional logic) the conjunction principle
(C) (OA & OB) ::::> O(A & B),
which cannot regarded as valid if the consistency principle is rejected:
perhaps a person can in some situation be subject to mutually incom-
patible obligations, but self-contradictory obligations are clearly impossi-
ble. (The standard system does not distinguish the consistency principle
from the principle that ought implies can, which denies the existence of
impossible obligations.) Neither (D) nor (C) belongs to what Brian F.
Chellas calls 'the minimal deontic logic' ([10], p. 202). Bas C. van Fraassen
[26] and P. K. Schotch and R. E. Jennings [22] have presented s~mantical
analyses of ought-statements which do not entail the validity of (D) or (C),
and which thus allow the possibility of genuine moral conflicts. The
problem of the resolution of nonnative conflicts has been investigated by
Carlos Alchourron and David Makinson [2], who show how a partial
ordering of a system of regulations may be used to resolve the inconsis-
tencies within the system.
In the standard approach to deontic logic, the concept permission
(permissibility) is defined simply as the absence of prohibition. The feature
ofthe standard system has been criticized on several grounds; for example,
it has been argued that the standard definition of permissibility excludes
the possibility of incomplete or open normative systems, that is, systems
which leave the normative status of some acts or states of affairs completely
undetermined. (Thus the standard approach cannot explain the distinction
between 'closed' and 'open' systems; cf. this volume, p. 166; von Wright
[27], Chapter IV; [28], pp. 413-415; and Carlos Alchourron and Eugenio
Bulygin [1], pp. 116-144, for discussions ofthis issue.) Perhaps the most
striking argument against the standard analysis of permissibility is pro-
vided by the 'paradox offree choice permission' or the 'paradox of disjunc-
tive permission' (see pp. 21 and 160 of this volume). Usually a disjunctive
permission seems to entail the permissibility of both disjuncts, but such an
inference is not justified by the standard system, in which 'PA' entails 'P(A
V B)" but not conversely. (Here 'P' is the permission operator.) Hans Kamp
[14] has presented an illuminating analysis of this problem in which he
shows that it is related to the peiformative use of permission sentences, that
is, the use of permission sentences for making previously prohibited acts or
states of affairs permissible and thus changing a normative system. In the
INTRODUCTION TO THE SECOND IMPRESSION xv
light of Kamp's discussion, the problem of disjunctive permission seems

more serious and more interesting that is suggested by Dagfinn F~llesdal
and Risto Hilpinen in their contribution to the present volume (pp. 22-23).
(This problem is also discussed by Hans Kamp in [15].) In [13], Risto
Hilpinen has presented an improved analysis of Georg Henrik von
Wright's concept of strong permission (cf. the present volume, pp. 160, 164--
165), and discussed the relationship of the problem of disjunctive per-
mission to an analogous problem in the logic of subjunctive conditionals
('the problem of disjunctive antecedents'; cf. also Barry Loewer [20]).
Normative concepts are usually applied to human actions: they belong
to practical discourse, and the logic of action concepts should therefore
form an essential part of the logical study of normative statements. The
standard approach to deontic logic has been criticized in this respect: for
example, Hector-Neri Castaneda ([5], p. 675) has argued that the standard
possible worlds semantics of deontic logic developed by Jaakko Hintikka,
Stig Kanger, and others is riot a satisfactory theory ofthe practical concept
of ought (or the concept of 'ought-to-do'; 'Tunsollen'), but concerns only
the concept of 'ought-to-be' ('Seinsollen'). Castaneda's own approach to
deontic logic is based on a fundamental semantic distinction between
practitions and propositions, or the distinction between actions practically
(or deontically) considered and the circumstances of actions (including
actions considered as circumstances or conditions of other actions), and he
has argued that this distinction is essential for a satisfactory solution of
various deontic paradoxes (cf. Castaneda [6], Chapter 7, and [7]).
However, Castaneda seems to have underestimated the interpretational
flexibility of the standard approach. In [11] Risto Hilpinen has tried to
show how the standard (possible worlds) semantics of deontic logic can
be interpreted as a theory of the practical concept of ought, and Jaakko
Hintikka's contribution to the present volume is clearly concerned with the
logic of Thnsollen: the individual variables of his system of quantificational
deontic logic are interpreted as variables for individual acts (pp. 60-67). A
satisfactory comprehensive theory of cormative concepts and action con-
cepts remains nevertheless still very much of a desideratum. Georg Henrik
von Wright has taken interesting new steps towards such a theory in his
recent paper [29], in which he discusses different ways of applying nor-
mative concepts to actions.
RISTO HILPINEN
xvi INTRODUCTION TO THE SECOND IMPRESSION
BIBLIOGRAPHY
[1] Alchourron Carlos E. and Bulygin, Eugenio, Normative Systems, Springer Verlag, Wien
and New York, 1971.
[2] Alchourron, Carlos E. and Makinson, David, 'Hierarchies of Regulations and their
Logic', New Studies in Deontic Logic, (ed. by Risto Hilpinen), D. Reidel Publ. Co.,
Dordrecht, 1981.
[3] al-Hibri, Azizah, Deontic Logic: A Comprehensive Appraisal and a New Proposal,
University Press of America, Washington, D.C., 1978.
[4] Aqvist, Lennart, and Hoepelman, Jaap, 'Some Theorems about a "Tree" System of
Deontic Logic', in New Studies in Deontic Logic, (ed. by Risto Hilpinen), D. Reidel Publ.
Co., Dordrecht, 1981.
[5] Castaneda, Hector-Neri, 'On the Semantics of the Ought-to-Do', in Semantics of Natural
Language, (ed. by Donald Davidson and Gilbert Harman), D. Reidel Publ. Co., Dor-
drecht, 1972, pp. 675-694.
[6] Castaneda, Hector-Neri, Thinking and Doing, D. Reidel Publ. Co., Dordrecht, 1975.
[7] Castaneda, Hector-Neri, 'The Paradoxes of Deontic Logic: The Simplest Solution to All
of Them in One Fell Swoop', in New Studies in Deontic Logic, (ed. by Risto Hilpinen), D.
Reidel Publ. Co., Dordrecht 1981.
[8] Chellas, Brian F., 'Imperatives', Theoria 37 (1971),114--129.
[9] Chellas, Brian F., 'Conditional Obligation', in Logical Theory and Semantic Analysis:
Essays Dedicated to Stig Kanger on His Fiftieth Birthday, (ed. by Soren Stenlund et al.),
D. Reidel Publ. Co., Dordrecht, 1974, pp. 23-34.
[10] Chellas, Brian F., Modal Logic: An Introduction, Cambridge University Press, Cam-
bridge, 1980.
[11] Greenspan, Patricia, 'Conditional Oughts and Hypothetical Imperatives', Journal of
Philosophy 12 (1975),259-276.
[12] Hilpinen, Risto, 'Deontic Logic and the Semantics of Possible Worlds', in Deontische
Logik und Semantik, (ed. by A. Conte, R. Hilpinen, and G. H. von Wright), Athenaion,
Wiesbaden, 1977, pp. 82-88.
[13] Hilpinen, Risto, 'Disjunctive Permissions and Conditionals with Disjunctive Antece-
dents', in Intensional Logic and Natural Language: Proceedings of the Second Soviet-
Finnish Logic Conference, Moscow 1979, (ed. by Ilkka Niiniluoto and Esa Saarinen),
Acta Philosophica F ennica, Helsinki, 1981.
[14] Kamp, Hans, 'Free Choice Permission', Aristotelian Society Proceedings N.S. 74 (1973-
74),57-74.
[15] Kamp, Hans, 'Semantics versus PragMatics', in Formal Semantics and Pragmatics for
Natural Languages, (ed. by F. Guenther and S. J. Schmidt), D. Reidel Publ. Co.,
Dordrecht 1979, pp. 255-287.
[16] Knuuttila, Simo, 'The Emergence of Deontic Logic in the Fourteenth Century', New
Studies in Deontic Logic, (ed. by Risto Hilpinen), D. Reidel Publ. Co., Dordrecht 1981.
[17] Lemmon, E. J. and Scott, Dana, The 'Lemmon Notes': An Introduction to Modal Logic,
(ed. by Krister Segerberg), American Philosophical Quarterly Monograph No 11, Basil
Blackwell, Oxford 1977.
[18] Lewis, David, Counterfactuals, Basil Blackwell, Oxford, 1973.
[19] Lewis, David, 'Semantic Analyses for Dyadic Deontic Logic', in Logical Theory and
Semantic Analysis: Essays Dedicated to Stig Kanger on His Fiftieth Birthday, D. Reidel
Publ. Co., Dordrecht 1974, pp. 1-14.
INTRODUCTION TO THE SECOND IMPRESSION xvii
[20] Loewer, Barry, 'Counterfactuals with Disjunctive Antecedents', Journal of Philosophy

73 (1976),531-537.
[21] Mott, Peter, 'On Chisholm's Paradox', Journal of Philosophical Logic 2 (1973),197-211.
[22] Schotch, Peter K. and Jennings, Raymond E., 'Non-Kripkean Deontic Logic', in New
Studies in Deontic Logic, (ed. by Risto Hilpinen), D. Reidel Pub!. Co., Dordrecht 1981.
[23] Thomason, Richmond, H., 'Deontic Logic as Founded on Tense Logic', in New Studies
in Deontic Logic, (ed. by Risto Hilpinen), D. Reidel Pub!. Co., Dordrecht, 1981.
[24] Thomason, Richmond H., 'Deontic Logic and the Role of Freedom in Moral Delibera-
tion', in New Studies in Deontic Logic, (ed. by Risto Hilpinen), D. Reidel Pub!. Co.,
Dordrecht 1981.
[25] van Fraassen, Bas C, The Logic of Conditional Obligation', Journal of Philosophical
Logic 1 (1972),417-438. Also in Exact Philosophy: Problems, Tools, and Goals, (ed. by
Mario Bunge), D. Reidel Pub!. Co., Dordrecht, 1973, pp. 151-172.
[26] van Fraassen, Bas C, 'Values and the Heart's Command', Journal of Philosophy 70
(1973),5-19.
[27] von Wright, Georg Henrik, An Essay in Deontic Logic and the General Theory of Action,
Acta Philosophica Fennica 21, North-Holland Pub!. Co., Amsterdam, 1968.
[28] von Wright, Georg Henrik, 'Problems and Prospects of Deontic Logic: A Survey', in
Modern Logic - A Survey: Historical, Philosophical, and Mathematical Aspects of
Modern Logic and Its Applications, (ed. by Evandro Agazzi), D. Reidel Pub!. Co.,
Dordrecht 1981, pp. 399-423.
[29] von Wright, Georg Henrik, 'On the Logic of Norms and Actions', in New Studies in
Deontic Logic, (ed. by Risto Hilpinen), D. Reidel Pub!. Co., Dordrecht, 1981.
DAGFINN F0LLESDAL AND RISTO HILPINEN
DEONTIC LOGIC: AN INTRODUCTION·
I. THE SUBJECT-MATTER OF DEONTIC LOGIC
The word 'deontic' is derived from the Greek word '3e6VtOl<;', which may
be translated 'as it should be' or 'duly'. Bentham uses 'deontology' for
"the science of morality", and Ernst Mally [30] was the first to use the
term - in the form Deontik - to refer to logical study of the normative use
of language. In accordance with Bolzano and Quine's definition of logical
truth, deontic logic can be defined as the study of those sentenc:es in
which only loaical words and normative expressions occur essentially.1
Normative expressions include the words 'obliption', 'duty', 'permission',
'right', and related expressions. These expressions may be termed Montic
words, and sentences involving them deontic sentences. a A deontic sen-
tence is a truth of deontic 1000c if it is true and remains true for all varia-
tions of its non-logical and non-deontic words (that is, expressions which
are not logical or deontic words). Deontic logic is closely related to the
logic of imperatives (or the logic of commands); in fact, many authors
regard these fields as essentially the same. 3 What is here called deontic
logic has also been referred to as logic of obligation and logic of norms
(or logic of normative systems).4
II. MALLY'S SYSTEM OF 'DEONTIK'
Most of the contemporary interest in deontic logic has been stimulated

by G. H. von Wright's classic paper 'Deontic Logic' ([46]; 1951). The
history of deontic logic goes farther back, however. The first philosopher
who attempted to build a formal theory of normative concepts was Ernst
Mally, a student of Alexius Meinong. In his monograph Grundgesetze
des Sol/ens: Elemente der Logik des Willens ([30]; 1926), Mally presented
an axiom system for the notion of ought.
According to Mally, judging (Urteilen) and willing (Wollen) are two
different attitudes towards states of affairs. Classical logic is the logic of
R. Hllpinell (ed.). Deolltic Logic: Introductory tmd SY8tellUltic Readlna8. 1-35. All right8 re8erved.
Copyright IC> 1970 by D. Reidel Publl8hillg ColfllHlllY. Dordrecht·Holll1llll.
2 DAGFINN FeLLESDAL AND RISTO HILPINEN
judgment; it lays down the criteria of correct and incorrect judgment.

Mally proposes to construct a similar logic for the attitude of willing that
a given state of affairs be the case ([30], pp. 1-2). This theory is termed
'Deontik'. A person's willing that a given state of affairs p be the case
may be expressed by sentences of the form 'p ought to be (the case)' (p
soil sein). This notion of ought is the deontic primitive of Mally's system.
The vocabulary of Mally's system consists of sentential letters (which
refer to (possible) states of affairs), sentential connectives, quantifiers
(ranging over propositions), and a sentential operator (which represents
the notion of oUght). Here (and throughout the present paper), we shall
employ the symbols 'p', 'q', 'r', ... as sentential symbols. We adopt the
customary conventions regarding the 'combining force' of various
sentential connectives, that is, '",' is the strongest connective, '&' is
stronger than' v', 'v' than' =>', and' =>' is stronger than' =='. 'p ought
to be the case' is expressed by 'Op', where '0' is a sentential operator.
In addition to the symbols mentioned above, Mally's system includes
four propositional constants, 'u', 'n', Ow' and 'm'.s 'u' stands for what is
'unconditionally obligatory', On' represents a state of affairs incompatible
with u, ow' stands for what is the case (i.e., for 'facts'), and om' represents
what is not the case. Thus n== '" u and m== '" w. Mally reads 'p => Oq' as
'p requires q' (p Jordert q), and expresses it briefly as 'pJq.' This abbre-
viation will not be used here.
Mally accepts the following Grundsiitze or axioms:
(AI) (p => Oq) & (q => r) => (p => Or)

(A2) (p => Oq) & (p => Or) => (p => O(q & r»
(A3) (p => Oq) == 0 (p => q)
(A4') (Eu) Ou
(AS') ,... (u => On)
Axiom (A4') says that there exists a state of affairs u which ought to be
the case. Mally calls such a state of affairs 'the unconditionally obligatory'
or 'unconditionally required'. In Mally's system 'u' is, however, a propo-
sitional constant; thus, according to customary systems of quantifi-
cational logic where the quantifiers range over ordinary individuals,
DEONTIC LOGIC: AN INTRODUCTION 3
(A4') is not well-formed. Mally's interpretation of (A4') can be expressed

simply as
(A4) Ou
The symbol on' occurring in (AS') is also a constant; it refers to a state
of affairs incompatible with u. Thus (AS') can be written as
(AS) '" (u :;) 0 '" u).
Mally calls (AS) the principle of the consistency of willing: what is un-
conditionally obligatory must not require a state of affairs incompatible
with it.
According to Mally, cOp' is true if and only if p is required by every
state of affairs ([30), p. 28). Thus we may add the following principle to
Mally's axioms (AI)-(AS):
(A6) Opifandonlyif,foreveryq,q::::> Op.
From the axioms (AI)-(A6) Mally derives about fifty theorems con-
cerning the notion of ought. In these deductions Mally employs the
customary inference schemata of propositional logic, and a rule which
permits the replacement of a formula in the scope of the operator '0' by
any logically equivalent formula. 6 Below, we shall present a sample of
Mally's theorems and deductions. The theorems are here presented in a
quantifier-free notation, and the constants ow' and om' have been elimi-
nated by replacing them by sentential variables, together with suitable
contingent premisses in terms of these variables. (For instance, 'Ow' is
replaced by- 'p::::> Op', 'p::::> Ow' by 'q:;) (p:;) Oq)', etc.) The symbol 'PL'
stands for 'propositional logic' .
(1) (q :;) r) :;) «p :;) Oq) :;) (p ::::> Or» (from (AI) by PL)
(2) (p ::::> Oq) ::::> (r :;) (p :;) Or» (from (1) by PL)
(3) (p :;) O(q & r» & (q & r ::::> q)
:;) (p :;) Oq) & (p :;) Or) (from (AI) by PL)
(4) (p :;) O(q & r» ::::> (p :;) Oq) & (p ::::> Or) (from (3) by PL)
(S) p :;) O(q & r) == (p :;) Oq) & (p ::::> Or) (from (A2) and (4»
(6) Op & Oq == O(p & q) (from (S) by (A6»
(7) Op v Oq:;) O(pv q) (from(6)bYPL)
4 DAGFINN F0LLESDAL AND RISTO HILPINEN
Most of these consequences appear fairly plausible from the intuitive

point of view. For instance, (6) states that p &q ought to be true if and
only if p ought to be true and q ought to be true. This principle seems
quite reasonable, and in fact is accepted also in most recently constructed
systems of deontic logic. All theorems of Mally's system are not, how-
ever, equally natural. Mally himself classifies some of the theorems he
derives as 'strange' (,befremdlich'). For instance, theorem (2), which is
an immediate consequence of (AI), appears somewhat 'strange'. It says
that if p requires something (i.e., if p is true only if something ought to
be the case), then p is true only if every proposition which is true ought
to be true. Other seemingly unacceptable principles occur among the
following formulae:
(8) Op&(p::::l q)::::l Oq (from (AI) by (A6»)

(9) (p ::::l q) ::::l (Op ::::l Oq) (from (8) by PL)
(10) Op ::::l (q ::::l Oq) (from (9) by PL)
(10) states that, if something ought to be, whatever is the case ought to
be the case. But worse follows:
(11) Ou ::::l (q ::::l Oq) (from (10»

(12) q::::l Oq (from (A4) by (11»
(13) u (from (AS) by PL)
(14) (u ::::l Oq) ::::l Oq (from (13) by PL)
(15) (u ::::l Oq) == Oq (from (14) by PL)
(16) (U::::l Oq)&(q::::l '" u)::::l (U::::l 0 '" u) (from(Al)byPL)
(17) '" (u ::::l Oq) & (q ::::l '" u» (from (16) and (AS»
(18) ,..,. (Oq & (q ::::l ,..,. u» (from (17) and (15»
(19) Oq ::::l ' " (q ::::l ' " u) (from (18) by PL)
(20) Oq::::l q (from (19»
(21) Oq == q (from (12) and (20»
(12) states that whatever is the case, ought to be, and according to (20),
the converse implication is also valid. Theorem (21), the conjunction of
(12) and (20), states that p ought to be the case if and only if it is the case.
(21) expresses the equivalence of ought and is (cf. [30], pp. 25 and 34).
These theorems are strongly counter-intuitive. Mally himself observes
that (21) is undoubtedly the strangest one among the 'strange' theorems
of his system ([30], p. 25).
He does not, however, reject his system of deontic logic on the basis
of theorems such as (12), (20) and (21). He constructs a philosophical
theory by means of which he endeavors to show that these consequences
are not as unnatural as they seem to be. For instance, Mally says that
his principles modify the presystematic notion of ought in important
respects; the presystematic notion involves ambiguities which are un-
covered by his logic ([30], p. 25). Nonetheless, it should be obvious that
results such as (21) are fatal to Mally's theory. In the first place, they are
in conflict with what Mally himself says about the notion of ought (or
obligation). For instance, Mally points out that
(22) Op v Oq
and
(23) 0 (p v q)
are not equivalent: (22) implies (23), but not conversely (cf. theorem (7»
([30], p. 27). This, of course, is what we should expect, but not so in
Mally's system. For in that system, (21) implies that (22) and (23) are
equivalent. A more serious consequence of (21) is that it makes deontic
logic trivial; according to (21), deontic logic is reducible to (non-modal)
propositional logic. By virtue of (21), cOp' is replaceable by 'p' every-
where it occurs; consequently Mally's axioms (Al)-(A3) reduce to tautol-
ogies of propositional logic, and both (A4) and (AS) are equivalent to u.
Where did Mally go wrong? It may be suggested that the inadequacies
of his system are attributable to a failure to draw a distinction between
(logical) implication and if-then statements (conditionals). For instance,
as was pointed out above, theorem (2) looks somewhat 'strange'. It is
an immediate consequence of (AI); hence axiom (AI) is not acceptable.
However, (AI) becomes much more plausible if the second conjunct of
the antecedent ('q:::l r') is replaced by 'r is logically implied by q' (or 'r is a
logical consequence of q'). Mally reads 'p :::I q' as 'p implies q'. The word
'implies' is notoriously ambiguous; in ordinary language it usually means
some sort of strict implication, not material implication. For instance,

the principle that cOp' implies 'Oq', if 'p' implies 'q', sounds plausible
enough, provided that 'implies' refers to strict implication (e.g., logical
implication), not to material implication (cf. (9».7 Another source of
confusion has probably been Mally's use of 'pJq' as short for 'p:::> Oq'.
Mally reads 'pfq' as 'pfordert q', i.e., 'p requires q' or 'doingp commits
(a person) to q'. Given this interpretation, the formula
(24) '" (uJ '" u)
looks quite plausible from the intuitive point of view, but if 'pfq' is
regarded as equivalent to 'p:::> Oq', (24) is identical with (AS). 'u' is an
immediate consequence of (AS); thus (AS) is not an acceptable principle
of deontic logic. Hence 'p :::> Oq' is not an adequate formalization of our
intuitions regarding the notion p Jordert q. Mally'S confusion at this point
is strikingly illustrated by the rather surprising axiom (A3). This confusion
itself is perhaps not so surprising; in recent literature on deontic logic,
there has been considerable discussion of the notion of commitment, and
both 'p:::> Oq' and '0 (p :::> q)' have been proposed as formalizations of
this notion (cf. Section VIII).
III. MALLY'S SUCCESSORS
Mally's system was criticized by Karl Menger in 'A Logic of the Doubtful.
On Optative and Imperative Logic' ([32], 1939). According to Menger,
Mally's main error was to attempt to build deontic logic upon the basis
of classical two-valued propositional logic. (This diagnosis has not been
borne out by later work in the field.) Menger suggests that deontic logic
should be based on three-valued logic, which includes, in addition to
the customary truth-values true andfalse, a third value, doubtful. Menger
constructs such a 'logic of the doubtful', and builds his imperative and
optative logic (the logic of wish) upon this three-valued system. Menger
has attempted to clarify the problems of ethics by formal methods also
in his book Moral, Wille und Weltgestaltung ([31]; 1934).
In addition to Menger's paper, there were published in 1939 papers on
imperative logic, a subject closely related to deontic logic, by Albert
Hofstadter and J. C. C. McKinsey [21], and by Rose Rand [39].8 Hof-
stadter and McKinsey's imperative logic can be criticized in the same
way as Mally's deonticlogic: in their system cOp' and 'p' are equivalent (or,
in the terminology employed by the authors, !p and p are equivalent). In
'Zur Logik der Sollsatze' ([13]; 1939), Kurt Grelling presented a system
of deontic logic which includes the following axioms:
(B1) (p & q => r) => (p & Oq => Or)
(B2) (p => Oq) => (Op => Oq)
(B1) is essentially the same as Mally's axiom (AI), and leads to similar
unacceptable consequences. This was pointed out by Karl Reach in [40].
Reach observed that if we replace 'q' by '", p' and Or' by 'p', we obtain
from (BI)
(25) (p & '" p => p) => (p & 0 '" p => Op).
The antecedent of (25) is a logical truth; thus (BI) implies
(26) P & 0 '" p => 0 p ,
that is, if we do something we ought not to do, we ought to do it. This

principle is strongly counter-intuitive. (BI) implies other unacceptable
theorems, too. For instance, Mally's theorem (10) is a consequence of
(B1). (10) states that if there is something we ought to do, we ought to
do whatever we happen to do. This is obviously unacceptable. Here,
Grelling is simply repeating Mally's errors, and these errors are probably
due to similar confusions concerning the notion of implication. The
consequences (26) and (10) of Grelling's (B1) have been observed by
A. N. Prior, who calls (26) "principle of the Jait accompli" and (10) "the
principle of continuous moral rectitude" ([37], pp. 227-228). These
principles should not be provable in acceptable systems of deontic logic.
Mally's and Grelling's failure to construct a viable system of deontic
logic may seem to reinforce the view that it is impossible to define logical
relations among deontic or imperative sentences, and that deontic or
imperative logic, as distinguished from the classical logic of propositions,
is not possible at all. In the case of imperatives, this pessimistic view is
based upon the fact that imperatives are not true or false. The notions
of validity and consistency studied in logic presuppose that truth-values
can be assigned to sentences; imperatives are not true or false; thus, so
the argument runs, it is not possible to apply the notions of validity,
logical consequence, etc., to imperative sentences. Vie'Ns of this kind

have been expressed by Jergen Jergensen [24], Alf Ross [44], and others.
On the other hand, we possess fairly definite intuitions regarding the
consistency and inconsistency of deontic (normative) statements, and it
seems worthwhile to attempt to elucidate and systematize such intuitions.
In deontic logic this difficulty can be avoided by treating deontic sentences
as descriptive sentences. According to this descriptive interpretation,
deontic sentences (represented by the formulae of deontic logic) describe
what is regarded as permitted, obligatory, forbidden, etc., in some un-
specified normative system. According to this interpretation, principles
of deontic logic are conditions of consistency for normative systems.
This descriptive interpretation of deontic formulae has been recommended
by Erik Stenius [45] and Bengt Hansson [16], among others. In what
follows, we shall often resort to this interpretation.
IV. VON WRIGHT'S SYSTEM OF DEONTIC LOGIC
The publication of G. H. von Wright's classic paper 'Deontic Logic' [46]

in 1951 was an important step forward in the development of deontic
logic. In this paper, von Wright presented the first viable system of deontic
logic. Most of the discussion of deontic logic after 1951 has been stimu-
lated - directly or indirectly - by von Wright's article. 9
Von Wright's approach to deontic logic is based upon the observation
that there exists a significant analogy between the deontic notions
obligation (ought) and permission and the modal notions necessity and
possibility. Obligation and permission are related to each other in the
same way as necessity and possibility: a proposition is necessary if and
only if its negation is not possible, and similarly a state of affairs (or an
act) p is obligatory if and only if", p is not permitted. According to von
Wright, deontic logic is a branch of modal logic. He calls obligation and
permission deontic modalities (or modes); necessity and possibility are
termed alethic modalities.
The notion of permission is the deontic primitive of von Wright's
system. 'p is permitted' is expressed briefly as 'Pp'. The notion of obli-
gation is defined in terms of 'r by
Von Wright's system includes three 'principles' or axioms, which can be

formalized as follows:
(C2) Pp v P '" P
This axiom is called "the principle of permission"; it states that, for any
act p, either p or '" p is permitted. If p is not permitted, it is called for-
bidden; thus (C2) can also be expressed in the form: It is not the case
that both p and '" p are forbidden.
(C3) pep v q) == Pp v Pq
(C3) states that p v q is permitted if and only if p is permitted or q is
permitted. This principle is termed by von Wright "the principle of
deontic distribution". According to (C3), the operator P distributes over
disjunctions. Both (C2) and (C3) have alethic analogues: Either p or '" p
is possible (i.e., both p and. '" p cannot be impossible), and a disjunctive
proposition p v q is possible if and only if p is possible or q is possible. 1o
The analogy between deontic and alethic modalities breaks down in the
case of the ab esse ad posse principle for alethic modalities: if p is true,
it is possible, but a state of affairs (or an act) is not necessarily permitted,
if it is the case (or the act is performed). (In Mally's deontic logic the
analogue of the ab esse ad posse principle holds, and this is a good reason
for rejecting Mally's system.) In addition to (C2) and (C3), von Wright
accepts a third principle, termed "the principle of deontic contingency":
(C4) '0 (p v '" p)' and' '" pcp & '" p)' are not valid.
If von Wright's deontic logic is formulated as an axiomatic system, (C4)
is, of course, superfluous; '0 (p v '" p)' and '", P(p & '" p)' are not
derivable from (C2) and (C3). Moreover, von Wright adopts the custom-
ary rules of inference of propositional logic and the following rule of
extensionality:
(C5) If 'p' and 'q' are logically equivalent, 'Pp' and 'Pq' are logically
equivalent.
According to von Wright, deontic operators are to be prefixed to
names of acts, not to descriptions of states of affairs. A similar view was
adopted earlier by Kurt Grelling. In this respect von Wright's and
Grelling's systems differ from Mally's. Mally took the sentential symbols
'p', 'q', etc. as referring to states of affairs; in von Wright'~ and Grelling's
systems they must be interpreted differently. Von Wright observes that
the word 'act' is ambiguous; it may be used to refer to "act-qualifying
properties", that is, general characteristics of acts (for instance, theft),
but also to individual acts, e.g., individual thefts ([46], pp. 1-2). Acts in
the former sense may be termed generic acts; acts in the latter sense are
termed act-individuals (Cf. [49], p. 36). In von Wright's system, deontic
operators are prefixed to names of generic acts; thus the symbols 'p',
'q', ... used above must be interpreted as standing for names of ge-
neric acts or act-predicates. This interpretation is philosophically sig-
nificant, and it also has certain purely syntactical consequences. If
deontic operators are prefixed to names of acts, ,the iteration of oper-
ators is not permissible: formulae such as 'Pp' and 'Op' are not act-
predicates, and consequently, e.g., 'OOp=.Op' and 'OPp :::l Pp' are not
well-formed formulae. For the same reason, 'mixed formulae', that is,
formulae in which propositional connectives are used to combine deontic
and non-deontic components (e.g., 'p :::l Op'), are not accepted as well-
formed. According to von Wright, it is, however, meaningful to speak
of the negation-act of a given act and of conjunction-, disjunction-, im-
plication-, and equivalence-acts of two acts. Thus, propositional logic is
applicable to, analysis of the logical relationships between generic acts.
Von Wright's system of deontic modalities is a decidable theory. The
validity of deontic formulae can be determined by a truth-table method.
In view of the restrictions imposed upon well-formedness of deontic
formulae, all well-formed formulae of von Wright's system have the
form (here we assume that all occurrences of '0' have been replaced by
'", P '" ')
(27) F{PflJ Pf2, ... , Pfm) ,
where 'F' represents a truth-function, and fl' /2, ... , In are formulae
of propositional logic. Let Pl' P2' ... , p" be all sentential letters (i.e.,
atomic formulae) occurring in/l ,f2, ... ,fm. Let d; be the perfect disjunc-
tive normal form offj in terms of Pl' P2' ... , P., and let c~, c~, ... , ct
be
the conjunctive parts of d/. According to the principle of extensionality,
(27) is equivalent to
and by virtue of the principle of deontic distribution, (28) is equivalent to

(29) F((Pe! v ... v Pei.), ... , (PeT v ... v Pew).
The formulae Pe~ are called the P-eonstituents of (27). P-constituents are
logically independent of each other except that all P-constituents cannot
be false; this is excluded by the principle of permission. Given n atomic
formulae Pi> there exist 2" different P-constituents and 22 " -1 truth-value
distributions over P-constituents Pe~. According to (27)-(29), every
deontic formula is a truth-function of its P-constituents; thus the truth-
value of any deontic formula can be decided for each value assignment
to P-constituents (by truth-tables). If a deontic formula has the value true
for all possible (or permitted) truth-value distributions over P-consti-
tuents, it is valid (a truth of deontic logic).
The perfect disjunctive normal form of 'p &,... p' is empty; thus the
normal form of 'P(p &,... p)' includes no P-constituents. The assignment
of truth-values to formulae of this type is, in a sense, a matter of con-
vention. Von Wright adopts the view that 'P(p &,.." p)' is not logically
false (the principle of deontic contingency), though the contrary assump-
tion is in many respects more natural, and accepted by most subsequent
authors on deontic logic.
By the mdhod described above, it can be shown that (C2), (C3), and,
e.g., the following formulae are valid:
(30) O(p&q)=Op&Oq
(31) Op v Oq ::::> O(p v q)
(32) P(p & q) ::::> Pp & Pq .
It should be observed that the converses of (31) and (32) are not valid.
In this respect Mally's intuitions were in accord with von Wright's
deontic logic (though his formal system was not). Moreover, von Wright
mentions and discusses the following principles:
(33) Op & O(p ::::> q) ::::> Oq
(34) Pp & O(p ::::> q) ::::> Pq
(35) ,.." Pq & 0 (p ::::> q) ::::> ,.." Pp
(36) O(p::::> q v r) & ,.." Pq & ,.." Pr ::::> ,.." Pp
(37) ...., (0 (p v q) & ...., Pp & ...., Pq)

(38) Op & O(p & q ~ r) ~ O(q ~ r)
(39) 0(...., p ~ p) ~ Op.
Von Wright calls these laws "the laws of commitment". According to

von Wright, the notion of commitment (that is, 'the performance of p
commits a person to perform q') may be formalized as 'O(p ~ q)'. Of
these principles, (39) is entirely trivial; according to the principle of
extensionality, it can be simplified to
(40) Op ~ Op.
(33) can also be expressed in the form

(41) O(p ~ q) ~ (Op ~ Oq).
It is interesting to compare this principle with Mally's theorem (9). The

unacceptable theorem (10) is an immediate consequence of (9); thus (9)
is also unacceptable. In Section II, it was pointed out that many counter-
intuitive theorems of Mally's system become much more plausible, if
they are formulated in terms of strict implication instead of material im-
plication; the!lrem (9) is a case in point. (41) is another seemingly accept-
able modification of (9). O(p ~ q) may be termed deontic implication;l1
according to (41), we obtain from (9) a valid principle, if the expression
'implies' used by Mally in his reading of (9) is interpreted as deontic
implication instead of material implication.
Von Wright mentions in a footnote that Thomas Aquinas frequently
refers to the laws (36) and (37). Aquinas draws a distinction between a
man's being perplexus simpliciter and a man's being perplexus secundum
quid. The former is the case if the man, without having done anything
forbidden, is, as such, obliged to choose between forbidden alternatives.
Aquinas denies that a man can be perplexus simpliciter - and this is also
denied by (37). A man is perplexus secundum quid, if he is obliged to
choose between forbidden alternatives as a result of a previous forbidden
act. According to Aquinas, this case is possible.I 2 Aquinas's view is in
accord with (36); (36) says that an obligation to choose between forbidden
alternatives implies that the person in question has committed a forbidden
act.
v. THB STANDARD SYSTBM OF DBONTIC LOGIC
Most systems of deontic logic include the system described above as a

subsystem. Thus, it may be rightly said that von Wright's work consti-
tutes the foundation of modern deontic logic.
As was mentioned above, von Wright accepts what he calls "the
principle of deontic contingency": ',.., P(p &,.., p)' and 'O(p v ,.., p)' are
not logically true. According to (32), 'P(p &,.., p)' implies 'P(p)'; hence
everything is permitted if 'P(p &,.., p)' is true. H everything is permitted,
nothing is obligatory; thus, the principle of deontic contingency states
merely that, as far as formal logic is concerned, there need not be any-
thing that is obligatory. The denial of this principle {that is, 'O(p v"", p)')
excludes only those cases in which nothing whatsoever is obligatory, in
other words, it excludes empty normative systeJll6. In a sense, '0 (p v,.., p)'
does not exclude even this, .since an obligation of this form is an 'empty'
obligation, that is, an obligation that it is impossible not to fulfill. Thus
the denial of (C4) seems fairly innocuous from the intuitive point of
view. In recent discussion of deontic logic, many authors have rejected
(C4) and accepted the axiom
(C4') ,.., P(p & ,.., p).
The axioms (C2), (C3) and (C4'), together with definition (Cl) and rule
(CS), constitute a system which will be called below the standard system
of deontic logic. 13 The axioms of the standard system can be formulated
in terms of the operator 0 as follows:
(01) Op :::> ,.., 0 "'" P
(02) O(p & q) == Op & Oq
(03) O(p v ,.., p)
If (02) is replaced by (41), we obtain a system equivalent to (0IH03).

In von Wright's system, deontic operators are prefixed to act-predicates.
Thus, the system of propositional logic which constitutes the basis of
von Wright's system is not, strictly speaking, a logic of propositions, but
a logic of act-names. In this logic, the notion of truth-value is replaced
by the notion of performance-value: a proposition may be true or false,
and similarly an act can be performed or not performed.14 This inter-
pretation of the formalism of propositional logic involves certain diffi-

culties. For instance, what does it mean to say that 'p & q' has the per-
formance-value performed? 'p & q' is a complex act-predicate exemplified
by those individual acts which are of type p and q; 'p & q is performed'
must thus be taken to mean that a single act of type p & q is performed.
However, on this interpretation the principles of propositional logic do
not seem to hold; for instance, from 'p' and 'q' it is not possible to infer
'p & q'. (If an act of type p is performed and an act of type q is performed,
it does not follow that an act of type p & q is performed.) The logic of
act-predicates satisfies the principles of propositional logic only if its
application is restricted to one single act-individual at a time. I5 This
requirement imposes a heavy restriction on the application of deontic
logic. In von Wright's system, deontic operators are prefixed to what can
be regarded, from a logical point of view, as names. Now it is natural to
inquire why we cannot introduce variables instead of the names and
quantifiers binding these variables, and thus attach deontic operators to
closed and open sentences. In fact, Jaakko Hintikka [19] has argued that
this generalization is not only natural, but also necessary if we want to
formalize certain deontic notions in an adequate way. In the same way,
it has been argued that in propositional deontic logic the symbols 'p',
'q', 'r', ... should be taken as representing closed sentences. These sen-
tences can be descriptions of states of affairs or statements about act-
individuals.
Most authors on deontic logic have accepted this reinterpretation of
von Wright's calculus. Erik Stenius [45] has interpreted the sentential
letters employed in the symbolism of deontic logic as descriptions of
act-individuals, and A. R. Anderson has interpreted them as descriptions
of possible states of affairs [1]. Also G. H. von Wright has recently
adopted this interpretation. IS It may be mentioned that this interpretation
is similar to Mally's interpretation of deontic logic. I7
If sentential letters 'p', 'q', etc. stand for sentences, it is no more
necessary to exclude 'mixed' formulae and formulae involving iterated
modalities, and consequently the question arises whether any formulae
of this type can be accepted as logical truths. For instance, the formulae
(42) Op::::> OOp
(43) OOp::::> Op
(44) O(Op ::J p)

are plausible-looking candidates for logical truth. (42)-{44) are not
derivable from (01)-(03). According to A. N. Prior, (44) is a valid
principle, and should be added to von Wright's system. (44) says that
what-ought-to-be ought to be; this principle seems very plausible from
the intuitive point of view (it must be distinguished from the trivial
principle that what ought-to-be, ought-to-be). (44) implies (43), but not
(42). In [37], p. 225 Prior also suggested that
(45) Op ::J «p ::J Oq) ::J Oq)
is valid, and should be subjoined to von Wright's axioms (it is not
derivable from (01)-{03». It can be shown, however, that (45) is not
acceptable; Prior's intuitions here seem to rest on confusions similar to
those discussed above in Sections II and 1I1.l8
VI. SEMANTICAL APPROACHES TO DEONTIC LOGIC
Most of the work described above represents the syntactical or axiomatic

approach to deontic logic. In this approach, plausible-looking candidates
for logical truth are selected as axioms for the notions studied, a sample
of consequences (or theorems) is derived from these axioms, and the
adequacy of the axiom system is decided by considering these conse-
quences. Oeontic formulae are normally interpreted simply by trans-
lating them to sentences of ordinary language. The plausibility of putative
theorems is judged on the basis of the intuitive plausibility of their
ordinary-language counterparts.
Mally's and Grelling's attempts to build systems of deontic and imper-
ative logic illustrate the dificulties involved in this approach. The
formulation of our intuitions concerning deontic notions in ordinary
language often involves ambiguous expressions such as 'implies', 're-
quires', etc. In many cases it is difficult to see what are the exact formal
counterparts of these intuitions, that is, what our intuitions really pertain
to. Mally's first axiom is a case in point. Moreover, a 'literal' translation
of formulae such as Op ::J OOp, POp, etc., to ordinary language yields
sentences which are hardly ever used at all. It is almost impossible to
decide whether such sentences are acceptable as principles of deontic
logic or not.
Recent studies by Stig Kanger [25], Saul A. Kripke ([271, [28]), laakko
Hintikka ([19], [20]), Richard Montague ([33], [34]), W. H. Hanson [15],
and others indicate that the application of modern semantical methods
(model theory) to deontic logic gives a more fruitful basis for understand-
ing deontic formulae and judging their acceptability. These semantical
theories concern quantified deontic logic, that is, deontic extensions of
first-order functional logic. Kanger's and Hintikka's theories are pre-
sented in detail elsewhere in the present volume; here we shall discuss
only the basic ideas of a semantics for propositional deontic logic. These
basic ideas are common to Hintikka's, Kanger's and Kripke's semantics.
The main difference between Hintikka's theory and the theories presented
by Kanger and Kripke is that the basic semantical device of Hintikka's
theory, a model set, is a set of formulae, whereas the models studied by
Kanger and Kripke are set-theoretical structures of a type more common
in modern semantics. The exposition below follows most closely Kripke
[27] and Hanson [15].
As was mentioned in Section III, deontic logic can be regarded as a
logical theory of normative systems. The principles of deontic logic
determine conditions of consistency for normative systems. By a 'norm-
ative system' we understand here simply any set of deontic sentences
closed under deduction.
When is a set of deontic sentences consistent? It seems natural to
require that at least the following 'minimal condition' should be satisfied:
(El) If a set of sentences A is consistent and {O 11> 012, ... ,0/.. , Pg}
s; A, then {fl' 12, ... , I .. , g} is consistent.
(El) says that a set of obligations is consistent only if all obligations in
this set can be simultaneously fulfilled, and that g is permitted only if it
can be realized without violating any of one's obligations. This seems
very plausible from the intuitive point of view. In fact, (El) (together
with rules for propositional connectives) is all we need for the standard
system of deontic logic: a deontic formula is provable in the standard
system if and only if its negation is inconsistent according to (El). It
should be observed that (El) does not require that all permitted states
of affairs can be realized simultaneously, but only that each permission
is compatible with all obligatory states of affairs. This shows that we are
here dealing with a fairly weak sense of 'permission': 'p is permitted'
means only that p is compatible with all obligatory states of affairs.19

A set of sentences is termed consistent (or satisfiable) if and only if
there is a possible state of affairs or a 'possible world' in which all mem-
bers of the set are true. In a more technical terminology, this can be
expressed by saying that a set of sentences is consistent if and only if
there is a model satisfying all members of the set. A model is binary
function yep, K), where 'p' is a variable ranging over atomic formulae,
and 'K' ranges over the elements of a given set S of 'possible worlds'. The
range of yep, K) is the set {T, F}, that is, V assigns to each atomic
formula (in the present case, to each sentential letter) a truth-value T or
F in each world KeS. Given a model, the assignment to truth-functional
compounds of atomic formulae is defined inductively: V(...., I. K)=T if
and only if V(1. K) =F; otherwise V(...., I. I) =F; V(J&g, K) =T if and
only if V(1. K)= V(g, I)=T; otherwise V(f&g, K)=F; V(fv g, K)=F
if and only if V(I.K)=V(g,K)=F; otherwise V(fvg,K)=T, etc. If
V(/;, K) =T for every f,eA, we say that A holds in K. According to (EI),
the following condition holds:
(Fl) If B = {all> 012, ... , Oln' Pg} holds in M, there is a world

Ml e S such that {fl' 12, ... , In> g} holds in Ml .
According to (FI), Ml is a world in which all those propositions are true

which ought to be true in M. Following Hintikka, we call worlds of this
kind deontic alternatives to M. We express the (deontic) alternativeness
of Ml to M by 'R(Mlo M)'. M may be thought of as representing our
actual world; deontic alternatives to Mare 'deonticallyperfect worlds' or
'ideal worlds' in which all obligations are fulfilled. All states of affairs
that hold in such worlds are permitted in our actual world. Now we can
replace (FI) by the following conditions:
(F2) If V(Pg, M) = T, there is a world M, e S such that

R(M;. M) and V(g, M,) = T.
(F3) If V(OI, M) = T, there is a world M, e S such that

R(M;. M) and V(f, M ,) = T.
(F4) If V(OI, M) = T, then V(f, M,) = T for every M, e S

such that R (Mb M),
1S DAGFINN F0LLESDAL AND RISTO HILPINEN
and define an assignment of truth-values to deontic formulae as follows:
(FS) V(O!, M) = T if and only if V{/, M i ) = T for each M, e S

such that R(M" M).
(F6) V(Pg, M) = T if and only if V(g, M,) = T for some M, e S
such that R (M" M).
(FS)~F6), together with rules for propositional connectives, determine

an assignment of truth-values for all formulae of the standard system.
The present semantical theory of deontic modalities can be represented
as an ordered triple [/ =(S, M, R), where S is a set of possible worlds,
MeS (M represents the 'actual world'), and R is a two-place relation
defined on S, called the relation of alternativeness. [/ is called a model
system (by Hintikka) or a model structure (by Kripke). A set of deontic
sentences is consistent (or satisfiable) if and only if there is a model
system [/ satisfying (Fl)~F4) such that all members of the set are true
in MeS. In terms of satisfiability, all the other basic semantical notions
(validity, contradictoriness, etc.) are easily defined. The standard system
of deontic logic is complete with respect to the present semantical theory:
f is valid according to this semantical theory if and only if it is a theorem
of the standard system.
It is possible to obtain stronger semantical systems of deontic logic by
imposing certain structural requirements on the relation R. For instance,
we may add to the conditions (F2)~F4) the condition
(F7) If R(M1' M) and R(M2' M 1), thenR(M2' M).
(F7) says that R is transitive. This condition has the same logical force
as the condition 20
(FS) If V (O!, M) = 1 and R(M!> M), then V (O!, M 1) = 1.
(FS) says that if an obligation holds in our actual world, it also holds in
its deontic alternatives. If this condition is accepted, (42) can be shown
to be valid. Moreover, it may be argued that ifMl is a deontically perfect
world (or an ideal world), then it should satisfy, not only such obligations
as hold in M, but also those which hold in Ml itself. In other words,
(F9) If there is a world MeS such that R(M1' M), then R(M1' Ml)'
Given this condition, (43) and (44) can be shown to be valid. 21 On the
other hand, it cannot be required that the relation R be reflexive in the
whole set S: R (M, M) is not an acceptable requirement, since our actual
world cannot be regarded as a deontically perfect world (on logical
grounds).
VII. REDUCTION OF DEONTIC LOGIC TO ALETHIC MODAL LOGIC
In [1] and [2], Alan Ross Anderson has suggested that deontic logic can
be reduced to alethic modal logic by means of a reduction schema
(G 1) 0 p == N (,..., p :::> S),
where 'S' is a propositional constant, and N is the modal necessity
operator. If it is assumed that the operator N satisfies, e.g., the axioms
of the Feys-von Wright modal system M,22 the axioms of the standard
system of deontic logic can be derived from (Gl) and the axiom
(G2) ,..., NS.
Anderson interprets'S' as a 'bad thing' or a sanction which results from
violation of one's duties. According to (Gl), p is obligatory if and only
if ,..., p (necessarily) implies the sanction S, in other words, p is forbidden
if and only if it implies the sanction. (G2) says that the sanction is avoid-
able, i.e., not everything is forbidden (or obligatory). According to (Gl),
the notion of permission (or permissibility) is defined by
(G3) Pp == M(p & ,..., S),
where M is the alethic possibility operator (Mp=="'" N,.., p). Thus, a state
of affairs p is permissible if and only if it is compatible with the absence
of the penalty S.
In an unpublished paper written in 1950, Stig Kanger presented a
simpler reduction schema,
(H1) Op == N(Q :::> p),
where 'Q' is a propositional constant, interpreted by Kanger as 'what
morality prescribes' (see [25], p. S3 in the present volume). According
to (HI), p is obligatory if and only if it is entailed by what morality pre-
scribes. Kanger's and Anderson's reduction schemata are closely related
to each other: If '8' is equivalent to',.." Q', (Gl) and (HI) are equivalent.
Given (H 1), the axioms of the standard system of deontic logic are deriv-
able from principles of alethic modal logic and the axiom
(H2) MQ,
which says that morality cannot require impossible states of affairs. The
notion of permission can now be defined by
(H3) Pp == M(Q & p).
Thus, p is permissible if and only if it is compatible (or 'compossible')
with 'what morality prescribes', i.e., with all obligatory states of affairs.
It has been argued that the notion of permission defined by (G3) is too
weak. (G3) says only that it is possible to do the permitted thing and
escape punishment. "Must it not, however, be as certain that the man
who does the permitted is not punished for what he has done as it is that
he who neglects his duty is punished?" (von Wright [54], p. 90). This
criticism is beside the point, however: the 'sanction' 8 is not relative to
specific acts, but results from the performance of any forbidden act. If
a person performs a permitted act p, then surely he should not be punished
for p, but he may at the same time perform another act which is forbidden,
and thus implies 8. Performance of a permitted act cannot guarantee an
escape from 8, but it preserves the possibility of such an escape.
Anderson's reduction schema may also be criticized on the grounds
that violation of one's duties does not invariably lead to punishment (cf.
von Wright [54], p. 90). From a purely formal point of view (Gl) is
quite unobjectionable; the above criticism does not concern the schema
(Gl) as such, but rather the interpretation of '8' as a penalty. 8 cannot
be a 'naturalistic' description of some actual penalty. G. H. von Wright
has suggested that '8' can be interpreted as liability (as opposed to
immunity) to punishment ([54], p. 93). Ifliability to punishment is thought
of as being involved in any violation of obligations on purely conceptual
grounds, this interpretation is, indeed, attractive. It may also be assumed
that 8 means only that the obligations included in a given normative
ct>de are not fulfilled, i.e., that deontically perfect circumstances are not
realized (cf. Anderson [3], pp. 345-347, and [4], p. 111; Prior [36], p. 146).
This interpretation of 8 seems to come closest to Kanger's interpretation
of Q as 'what morality prescribes'. These interpretations of (Gl) and
(HI) are unproblematic, if 'Q' ('what morality prescribes') and'S'

(realization of forbidden circumstances) are defined in terms of some
specific normative system T, and 'Of' means that/is obligatory according
to T. In this case the notion of necessity involved in (GI) and (HI) can
be regarded as analytic or logical necessity.23
The basic idea underlying Kanger's reduction schema has been applied
in an interesting way to the analysis of certain deontic logics by Bengt
Hansson ([16]; reprinted in this volume). Essentially the same idea has
been applied to inductive logic and the logic of rational belief by Isaac
Levi [29] and Risto Hilpinen [17]. 24 Whether or not Kanger's and Ander-
son's schemata can be regarded as genuine 'reductions' of deontic logic
to alethic modal logic, they provide an interesting insight into the
structure of deontic logics, and have also led to fruitful applications. 25
VIII. PARADOXES IN DEONTIC LOGIC
The standard system has been critized on the grounds that it includes
'paradoxical' theorems. For instance, many philosophers have felt that
there is something paradoxical in the formula
(46) Op =:l O(p v q),
which is a theorem of the standard system. This theorem says that if a
certain state of affairs p ought to be the case, then also p v q ought to be
the case. For example, if I ought to mail a letter, I also ought to mail or
burn it. But if I in fact ought to mail a letter, then surely it is awkward to
say that I ought to mail or burn it. This example has been presented by
Alf Ross [44], and the paradox involved in it (and other similar examples)
is called Ross's paradox. A similar paradox can be formulated in terms
of the notion of permission as follows:
(47) Pp ::> P(p V q)
is a theorem of the standard system. According to (47), if a person is
permitted to smoke, he is also permitted to smoke or kill. Now a per-
mission to smoke may sound innocuous enough, but its alleged conse-
quence, a permission to smoke or kill, seems decidedly immoral.
These applications of (46) and (47) may indeed sound paradoxical,
but they lose most of their paradoxical character as soon as we pay
sufficient attention to the proper interpretation of the standard system.

In terms of the semantics sketched in Section VI, (46) may be interpreted
as saying that if p is true in all deontically perfect worlds, p v q is also
true in all deontically perfect worlds. This is no more paradoxical than
the fact that p v q is a logical consequence of p. In the case of Ross's
example, it may be more appropriate to speak of deontically perfect
sequences of events than of deontically perfect worlds. It should be clear
that if every deontically perfect sequence of events satisfies the description
'a mails a letter', the description 'a mails a letter or burns it' is also
satisfied by such sequences of events. According to the standard system
of deontic logic, (47) means that if p is logically compatible with a
person's obligations, p v q is also compatible with them; thus (47) is as
unparadoxical as (46).
Nevertheless, some sort of explanation is needed for the seemingly
paradoxical character of the above examples. However, it is possible
that such an explanation has nothing to do with deontic logic in particular.
The paradoxes mentioned above may perhaps be explained by reference
to very general conventions regarding the use of language. For instance,
it is generally assumed that a person makes as strong statements as he
is in a position to make. If someone wants another person to mail a
letter, it is surely very odd for him to say that the letter ought to be
mailed or burned, especially if the latter alternative is forbidden. Similar
remarks apply to (47). If we want to explain the actual uses of deontic
expressions in ordinary language, such general conventions must be
taken into account, but they need not be incorporated into deontic logic. 26
In the case of the 'paradox' involved in (47), it is important to observe
that in ordinary language, the logical force of the word 'or' is in some
cases the same as that of 'and'. For instance, in many cases the sentence
'a may do p or q' is used to express the same statement as 'a may do p
and a may do q'. This fact has led some philosophers to assume that these
cases involve a special notion of permission, termed free choice permission.
G. H. von Wright ([52], pp. 21-22) has suggested that a free choice
permission and the permission concept defined by the standard system
of deontic logic have different logics; the former concept does not satisfy
the distribution principle (C3), but instead the law
(48) pep v q) == Pp & Pq.

The 'paradoxical' theorem (47) is not valid for this notion of permission.
According to von Wright, the notion of free choice permission cannot
be formalized in the standard system. It seems to us, however, that a free
choice permission can be expressed in the standard system in a perfectly
adequate way: Pp&Pq. If 'a is permitted to smoke or kill' is a free choice
permission, it should be formalized as Pp &Pq, and this is not, of course,
implied by Pp (according to the standard system). If the word 'or' is
interpreted in this way (as it often is in ordinary language), 'a is permitted
to smoke' does not imply 'a is permitted to smoke or kill'. There is no
need to invent special notions of permission or construct special logics
of permission and obligation on the basis of this accidental interchange-
ability of the words 'or' and 'and' in ordinary language. 27
All paradoxes discovered in the standard system of deontic logic are
not, however, as uninteresting as those discussed above. There is a group
of paradoxes which, so it seems to us, show something significant about
the limitations of the standard system. This group has variously been
called the paradoxes of derived obligation, paradoxes of commitment, or
paradoxes of contrarY-fo-duty imperatives. 28
As was mentioned above, von Wright [46] formalizes the notion of
commitment by formulae of the type
(49) O(p => q),
and reads (49) as 'performance of p commits a person to performance

of q'. This formalization of commitment has certain paradoxical con~e
quences, however, some of which have been pointed out by A. N. Prior
[35]. According to the standard system,
(50) 0 '" p => O(p => q)
is a valid formula. If (49) is regarded as a formalization of commitment,

(50) says that doing something forbidden commits one to anything. This
appears counter-intuitive. Of course, if properly interpreted, (50) is a
perfectly sound formula. It is equivalent to
(51) 0", P => 0(- p v q),
which has the same form as (46). Nevertheless, (50) suggests that (49) is
- at least in some cases - an inadequate formalization of commitment.
In Formal Logic ([37], pp. 224-225), Prior suggested that the notion of
commitment may be formalized by formulae of the type
(52) p ::l Oq.
This alternative was not available to von Wright in his 1951 system, since
(52) is not a well-formed formula in that system. However, if deontic
operators are prefixed to sentences, (52) may be accepted as well-formed.
If the notion of commitment is formalized as (52), the paradox corre-
sponding to (50) does not arise;
(53) 0 '" P ::l (p ::l Oq)
is not a theorem of the standard system (or the systems obtained by

addition of (42)-(44) to the standard system). However, also this formal-
ization has certain undesirable consequences. The formula
(54) '" P ::l (p ::l Oq)
is a tautology of propositional logic. (54) says that whatever is not done

(or is not the case) commits us to anything. Of course, (54) is quite un-
objectionable as such, but it raises doubts as to the present formalization
of commitment.
The first-mentioned 'paradox of commitment' arises in cases in which
something forbidden is done (or some forbidden state of affairs is the
case). A highly interesting example of this type has been presented by
Roderick M. Chisholm in [8]. Consider the following sentences:
(55) It ought to be that a certain man go to the assistance of his

neighbours.
(56) It ought to be that if he does go he tell them he is coming.
(57) If he does not go then he ought not to tell them he is coming.
(58) He does not go.
How should we formalize this example in the terminology of the standard

system of deontic logic? Let 'p' be short for 'the man goes to the assistance
of his neighbours', and let 'q' stand for 'the man tells his neighbours that
he comes to their assistance'. Now the sentence (60) can obviously be

expressed as
(59) Op,
and (56) seems to have the form

(60) O(p => q).
(57) says that if not p, then the man ought not to do q, that is,
(61) ,...., p=>O ,...., q;
and (58) is simply
(62) ,...., p
Chisholm points out that (59-62) imply a contradiction. According to the
standard system, (55) and (56) imply
(63) Oq,
whereas (57) and (58) imply (by Modus Ponens)
(64) 0 ,...., q.
By virtue of the axiom (Dl), (63) implies

(65) ,...., 0 ,...., q,
which contradicts (64). However, from the intuitive point of view, the
sentences (55H58) do not seem to involve a contradiction. (55H57)
seem to constitute a perfectly. reasonable set of regulations, and the
addition of the contingent premiss (58) to them should not make the
set inconsistent.
Now it may be suggested that there is an unjustified logical asymmetry
in the premisses (60) and (61): in (60) '0' precedes '::::>', but in (61) their
order is reversed. This asymmetry is to be found in (56) and (57), but it
may be argued that this is just a linguistic accident, and has no logical
importance. If (60) or (61) is modified in accordance with this proposal,
the contradiction is avoided, but then we get paradoxes which resemble
the paradoxes of commitment discussed above. 29
The notion of ought occurring in sentence (57) is termed by Chisholm
a contrary-to-duty imperative. A contrary-to-duty imperative says what a
person ought to do if he has violated his duties. Such imperatives (or

ought's) are widely used in ordinary language, and they are of importance
for ethical theories, "for most of us need a way of deciding, not only
what we ought to do, but also what we ought to do after we fail to do
some of the things we ought to do" (Chisholm [8], p. 36). The above
example shows - or at least strongly suggests - that such imperatives
cannot be formalized in the standard system of propositional deontic
logic. 30
In view of the semantical theory of deontic modalities presented in
Section VI, it is easy to see why contrary-to-duty imperatives cannot be
formalized in the standard system. According to this theory, a sentence
Of can be interpreted as saying that f is true in -all deontically perfect
worlds or in all ideal alternatives to our 'actual world'. The sentences
(55) and (56) are of this type. In all ideal worlds it is true that the man
of Chisholm's example goes to assist his neighbours, and also tells them
that he is coming. We shall express this by saying that in the standard
system sentences of the form 'Of' pertain to ideal worlds ('pertain' is
here a semi-technical term). Sentence (57) does not pertain to ideal
worlds in this sense. It can be taken to mean that . . . , q is true in all worlds
in which . . . , p is true, but which otherwise are as 'ideal' as worlds satis-
fying . . . , p can possibly be. Such 'almost ideal' worlds are not deontically
perfect, since . . . , p is false in every deontically perfect world. This notion
of ought cannot be expressed in the standard system, since the standard
system includes only one deontic operator which pertains to deontically
perfect worlds. The formalization of (57) requires another modal oper-
ator, say 0·, or 0 _ p (the latter notation is perhaps more appropriate,
as the obligation expressed by (57) seems to be relative to . . . , p). Of course,
it may be assumed that this new operator satisfies the axioms of the
standard system.
IX. SYSTEMS OF DYADIC DEONTIC MODALITIES
On the basis of the paradoxes of commitment, G. H. von Wright con-

cluded that the notion of commitment (derived obligations) cannot be
formalized in the standard system in an adequate way. In [48] he proposed
a new system of deontic logic which was intended to capture the idea of
derived obligation. The deontic primitive of this system is a conditional
I
notion of permission, P(p r), which may be read as 'p is permissible
under circumstances r'. The system has two axioms:
(11) P(p! r) v P(,.." p! r)

(12) P(p&q!r)==P(p!r)&P(q!r&p),
and the same rules of inference as the standard system. The notion of
conditional obligation is defined by
(13) O(p! r) == ,.., P(,.., P I r).

Axioms (11)-(12) imply
(66) ! !
P (p v q r) == P (p r) v P (q r). I
I
According to (11) and (66), under constant circumstances pep r) and
I
O(p r) satisfy the axioms of von Wright's original (1951) monadic
system.
In [50] von Wright presented another system of dyadic deontic mod-
alities, which includes the following axioms:
(K2) I
,.., (O(p r) & 0(,.., pi r»)
(K3) 0 (p & q ! r) == 0 (p ! r) & 0 (q I r)
(K4) O(p I r v s) == O(p I r) & O(p ! s)
In the same way as in the case of the monadic system, we may exclude
'empty' normative systems by adding to (K2)-(K4) the axiom
(Kl) O(p v '" p! r).

The notion of conditional permission is defined in the customary way:
(K5) I
P (p r) == '" 0 ('" p I r).
(Kl )-(K3) say that under constant circumstances the notion of condi-
tional obligation satisfies the principles of the monadic standard system.
Axiom (K4) is more interesting; it is specific to conditional obligations.
In [51] von Wright mentions that the construction of this system was
stimulated by Chisholm's discussion of contrary-to-duty imperatives
(pp. 103-104; p. 115 in this volume). Below, we shall test the adequacy
of von Wright's system by applying it to Chisholm's example. 31
We assume that sentences (55) and (56) express 'absoll1te' obligations.

In von Wright's system, obligations of this type can be expressed by
I
formulae of the type 'O(p t)', where 't' stands for a tautologous for-
mula. Thus (55) and (56) can be formalized as
(67) O(p It)
and
(68) O(p::::> q It),
respectively. (57) expresses a contrary-to-duty imperative, that is, an
obligation which holds under circumstances '" p. It can now be formal-
ized as
(69) O( '" q I '" p).
(67) and (68) imply
(70) O(q It),
but its negation is not derivable from (69) and (62). However, Chisholm's
paradox follows from (67)-(70) in another way. Von Wright's axiom (K4)
implies
(71) O(plr)::::>O(plr&s);
that is, if p is obligatory under certain circumstances, it is also obligatory
under all logically stronger circumstances. In particular,
(72) O(p I t) ::::> O(p I r)
holds for any r. By virtue of (72), (70) implies
(73) O(q I '" p),
and according to (K2), this implies
(74) '" O( '" q I '" p),
which contradicts (69). A similar paradox has been pointed out to von
Wright by Peter Geach (see [51], p. 104).
The present difficulty seems to be due to von Wright's axiom (K4) (as
has been argued by Bengt Hansson in [16]), which permits the inference
from (70) to (73). Thus it seems natural to reject (K4). In [51] von Wright
does not do this, however; instead he rejects (K2). (K2) is equivalent to

(75) I
O(p r)::> - 0(- pi r).
If (75) is rejected, (69) and (73) are no longer formally inconsistent. Von
Wright restricts the application of (K2) to absolute obligations; he
replaces it by the weaker axiom
(K2') I
- (O(p t) & O( - pit»).
According to von Wright, normative systems which include jointly in-
consistent absolute obligations are logically unacceptable. However, in
his view it is possible that logically acceptable normative systems imply
inconsistent conditional obligations. (By 'inconsistent' obligations we
mean obligations which do not satisfy (K2).) Von Wright calls a situ-
ation which creates such inconsistent obligations a predicament.
It seems to us, however, that this solution simply does not work in
the case of Chisholm's example. According to von Wright's logic,
Chisholm's example involves a predicament: both q and ,.., q are obli-
gatory in the circumstance ,.., p. Moreover, (K3) implies that q&,.., q is
also obligatory given ,.., p, and hence everything is obligatory. But this
simply is not the case. Chisholm's point is that we need a way of deciding
what we ought to do after we fail to do something we ought to do (in the
first place). This is precisely what von Wright's system does not accom-
plish.
Moreover, it seems to us that normative systems which (together with
contingent factual premisses) may imply a predicament cannot be accepted
as consistent and logically acceptable. According to von Wright, a
predicament arises, e.g., if a person promist;s to do something forbidden.
All promises ought to be kept; therefore the promise commits the person
to do something forbidden. This is a predicament. We would take this
example as evidence that normative systems according to which all
promises (independently of the nature of the promise) ought to be kept,
are logically objectionable. Thus it seems to us that the axiom (K2)
should be preserved in its original form, and (K4) must be rejected. 32
Axiom (K4) is equivalent to a conjunction of two formulae,
(K4.1) I
O(p r) & O(p s) I ::> I v s)
O(p r
and
(K4.2) I
O(p r v s) ::> I
O(p r) & O(p Is).
The counter-intuitive consequences of (K4) mentioned abl)ve follow from

(K4.2). In Section VIII, it was suggested that a contrary-to-duty imper-
ative of the form 'if r, then it ought to be the case that p' or 'if r, then a
ought to do p' may be interpreted as follows: p is true in all worlds in
which r is true but which otherwise are as ideal as worlds satisfying r can
possibly be. This interpretation can also be expressed as follows:
(Ll) I
O(p r) means thatp is true in all (possible) worlds in which
r is true, and which resemble deontically perfect worlds in
other respects as much as possible.
Let us call worlds in which r is true, and which resemble in other respects
deontically perfect worlds as much as possible, r-ideal worlds. Thus, (Ll)
says that 0 (p I r) is true if and only if p is true in all r-ideal worlds;
hence P (p I r) means that p is true in some r-ideal world. If r is a logical
truth, r-ideal worlds are, of course, identical with deontically perfect
worlds.
According to (Ll), (K4.1) says that if p is true in all r-ideal worlds
and in all s-ideal worlds, it is true also in all r v s-ideal worlds. r v s is
true in all r v s-ideal worlds; hence either r or s is true in every r v s-ideal
world. No r v s-ideal world can resemble deontically perfe.ct worlds
more than ev~ry r-ideal and every s-ideal world: if some r v s-ideal world
resembled deontically perfect worlds more than any other world in
which either r or s is true, it would be either an r-ideal or an s-ideal
world. Conversely, r v s-ideal worlds resemble deontically perfect worlds
at least as much as r-ideal or s-ideal worlds: if some r-ideal (or s-ideal)
world resembled deontically perfect worlds more than any other world
in which r v s is true, it would count as an r v s-ideal world. Conse-
quently r v s-ideal worlds are a subset of the union of r-ideal and s-ideal
worlds, and thus (K4.1) is valid. (K4.2) says that if p is true in all r v s-
ideal worlds, it is true in all r-ideal worlds and in all s-ideal worlds. This is
not always the case. For instance, if s= '" r, r v s is equivalent to t, and
(K4.2) implies that p is true in all r-ideal worlds, if it is true in all deontic-
ally perfect worlds. If r is false in every deontically perfect world, i.e., if
r is forbidden, p may be true in all deontically perfect worlds without
being true in any r-ideal world. Thus (K4.2) is not valid. On the other
hand, if r is permitted, it is true in some deontically perfect world, and
such a world is, of course, also an r-ideal world: in this case r-ideal
worlds are a subset of all deontically perfect worlds. Thus the following
modification of (72) is valid:
(76) O(p It) & P(r It) ::> O(p r).I
More generally, it can be shown that according to (Ll), (~4.2) must be
weakened to
(77) I
O(p r v s) & P(r s) I ::> I
O(p r).
Bengt Hansson has recently proposed for conditional deontic modalities
a semantical foundation of the type sketched above ([16]; see pp. 143-
146 in this volume). Hansson's semantical system combines the logic of
preference with deontic logic. This theory is in accord with Chisholm's
requirement: according to Hansson's theory, it is possible to give reason-
able answers to the question of what we ought to do after we have failed
to fulfill our 'absolute' obligations. It may be expected that semantical
theories of this type will eventually provide a solid intuitive foundation
for systems of conditional deontic modalities.
University of Oslo
University of Turku
BIBLlOGRAPHy33
[1] Anderson, Alan Ross, The Formal Analysis of Normati"e Systems (Technical
Report No.2, Contract No. SAR/Nonr-609 (16), Office of Naval Research,
Group Psychology Branch), New Haven 1956. Reprinted in The Logic of Decision
and Action (ed. by N. Rescher), University of Pittsburgh Press, Pittsburgh, 1967,
pp. 147-213.
[2] Anderson, Alan Ross, 'A Reduction of Deontic Logic to A1ethic Modal Logic',
Mind 67 (1958) 100-103.
[3] Anderson, Alan Ross, 'Some Nasty Problems in the Formal Logic of Ethics',
Nous 1 (1967) 345-360.
[4] Anderson, Alan Ross, 'Comments on von Wright's "Logic and Ontology of
Norms"', in Philosophical Logic (ed. by J. W. Davis, D. J. Hocmey, and W. K.
Wilson), D. Reidel Publ. Co., Dordrecht, 1969, pp. 108-113.
[5] Bar-Hillel, Yehoshua, 'Imperative Inference', Analysis 16 (1966) 79-82.
[6] Broad, C. D., 'Imperatives, Categorical &. Hypothetical', T71ePhilosopher 1 (1950)
62-75.
[7] Castafteda, Hector-Neri, 'The Logic of Obligation', Pliilosophical Studies 10 (1959)
17-23.
[8] Chisholm, Roderick M., 'Contrary-to-duty Imperatives and Deontic Logic',
Analysis 24 (1963) 33-36.
[9] Dawson, E. E., 'A Model for Deontic Logic', Anolysis l' (1959) 73-78.
[10] Feys, Robert, Modal Logics (ed. with some complements by J. Dopp), Gauthier-
Villars, Paris, 1965.
[11] Fisher, Mark, 'A Logical Theory of Commanding', Logique et Analyse 4 (1961)
154-169.
[12] Gombay, Andre, 'Imperative Inference and Disjunction', Analysis 2S (1965)
58-62.
[13] Grelling, Kurt, 'Zur Logik der Sollsiitze', Unity of Science Forum, January 1939,
44-47.
[14] Grice, H. P., 'The Causal Theory of Perception', Proceedings of the Aristotelian
Society, Suppl. Vol. 3S (1961); reprinted in Perceiving, Sensing and Knowing (ed.
by R. Swartz), Doubleday and Co., New York, 1965, pp. 438-472.
[15] Hanson, W. H., 'Semantics for Deontic Logic', Logique et Analyse 8 (1965)
177-190.
[16] Hansson, Bengt, 'An Analysis of Some Deontic Logics', Nous 4 (1970) 373-398.
Reprinted in this volume, pp. 121-147.
[17] Hilpinen, Risto, Rules of Acceptance and Inductive Logic (Acta Philosophica
Fennica 22), North-Holland Publ. Co., Amsterdam, 1968.
[18] Hilpinen, Risto, 'An Analysis of Relativised Modalities', in Philosophical Logic
(ed. by J. W. Davis, D. J. Hockney and W. K. Wilson), D. Reidel Publ. Co.,
Dordrecht, 1969, pp. 181-193.
[19] Hintikka, Jaakko, Quantifiers in Deontic Logic (Societas Scientiarum Fennica,
Commentationes Humanarum Litterarum 23:4), Helsinki 1957.
[20] Hintikka, Jaakko, 'Deontic Logic and Its Philosophical Morals', in Models for
Modalities. Selected Essays (by J. Hintikka), D. Reidel Publ. Co., Dordrecht,
1970, pp. 184-214.
[21] Hofstadter, Albert and McKinsey, J. C. C., 'On the Logic of Imperatives',
Philosophy of Science 6 (1939) 446-457.
[22] Hohfeld, Wesley Newcomb, Fundamental Legal Conceptions as Applied in Judical
Reasoning (ed. by W. W. Cook), Yale University Press, New Haven, 1919.
[23] Hughes, G. E. and Cresswell, M., An Introduction to Modal Logic, Methuen and
Co., London 1968.
[24] Jl!lrgensen, Jl!lrgen, 'Imperatives and Logic', Erkenntnis 7 (1937-8) 288-296.
[25] Kanger, Stig, New Foundations for Ethical Theory, Stockholm 1957. Reprinted in
this volume, pp. 36--58.
[26] Kanger, Stig and Kanger, Helle, 'Rights and Parliamentarism', Theoria 32 (1966)
85-115.
[27] Kripke, Saul Aaron, 'Semantical Analysis of Modal Logic I: Normal Modal
Propositional Calculi', Zeitschrift fur mathematische Logik und Grundlagen der
Mathematik 9 (1963) 67-96.
[28] Kripke, Saul Aaron, 'Semantical Considerations on Modal Logic', Proceedings
of a Colloquium on Modal and Many-Valued Logics (Acta Philosophica Fennica
16), Helsinki 1963, pp. 83-94.
[29] Levi, Isaac, Gambling with Truth, Alfred A. Knopf, New York, 1967.
[30] Mally, Ernst, Grundgesetze des Sol/ens. Elemente der Logik des Willens. Leuschner
& Lubensky, Graz, 1926.
[31] Menger, Karl, Moral, Wille und Weltgestaltung. Grundlegung der Logik der Sitten,
Wien 1934.
[32] Menger, Karl, 'A Logic of the Doubtful. On Optative and Imperative Logic',
Reports of a Mathematical Colloquium 1 (Notre Dame University, Indiana

University Press), 1939, pp. 53-64.
[33] Montague, Richard, 'Logical Necessity, Physical Necessity, Ethics, and Quanti-
fiers', Inquiry 4 (1960) 259-269.
[34] Montague, Richard, 'Pragmatics', in Contemporary Philosophy, Vol. I: Logic and
the Foundations of Mathematics (ed. by R. Kllbansky), La Nuova Italia Editrice,
Firenze, 1968, pp. 102-122.
[35] Prior, A. N., 'The Paradoxes of Derived Obligation', Mind 63 (1954) 64-65.
[36] Prior, A. N., 'Escapism: The Logical Basis of Ethics', in Essays in Moral Philos-
ophy (ed. by A. I. Melden), University of Washington Press, Seattle, 1958,
pp. 135-146.
[37] Prior, A. N., Formal Logic (2nd ed.), Oxford University Press, Oxford, 1962.
[38] Quine, W. V. 0., 'Carnap and Logical Truth', in The Philosophy of RudolfCarnap
(ed. by P. A. Schilpp), Open Court, La Salle, 111.,1963, pp. 385-406.
[39] Rand, Rose, 'Logik der Forderungssiitze', Revue internationale de la tMorie du
droit 1 (1939) 308-322.
[40] Reach, Karl, 'Some Comments on Grelling's Paper "Zur Logik der Sollsitze"',
Unity of Science Forum, April 1939, p. 72.
[41] Rescher, Nicholas, 'An Axiom System for Deontic Logic', Philosophical Studies
9 (1958) 24--30.
[42] Rescher, Nicholas, 'Conditional Permission in Deontic Logic', Philosophical
Studies 13 (1962) 1-6.
[43] Rescher, Nicholas and Robinson, John, 'Can One Infer Commands from Com-
mands?', Analysis 14 (1964) 176-179.
[44] Ross, Alf, 'Imperatives and Logic', Theoria 7 (1941) 53-71.
[45] Stenius, Erik, 'Principles of a Logic of Normative Systems', in Proceedings of a
Colloquium on Modal and Many- Valued Logics (Acta Phiwsophica Fennica 16),
Helsinki 1963, pp. 247-260.
[46] von Wright, Georg Henrik, 'Deontic Logic', Mind 60 (1951) 1-15. Reprinted in
Logical Studies (by G. H. von Wright), Routledge and Kegan Paul, London,
1957, pp. 58-74.
[47] von Wright, Georg Henrik, An Essay in Modal Logic, North-Holland Pub!. Co.,
Amsterdam, 1951.
[48] von Wright, Georg Henrik, 'A Note on Deontic Logic and Derived Obligation',
Mind 65 (1956) 507-509.
[49] von Wright, Georg Henrik, Norm and Action, Routledge and Kegan Paul,
London 1963.
[50] von Wright, Georg Henrik, 'A New System of Deontic Logic', Danish Yearbook
of Philosophy 1 (1964) 173-182. Reprinted in this volume, pp. 105-115.
[51] von Wright, Georg Henrik, 'A Correction to a New System of Deontic Logic',
Danish Yearbook of Philosophy 1 (1965) 103-107. Reprinted (in part) in this
volume, pp. 115-120.
[52] von Wright, Georg Henrik, An Essay in Deontic Logic and the General Theory of
Action (Acta Philosophica Fennica 11), North-Holland Pub!. Co., Amsterdam,
1968.
[53] von Wright, Georg Henrik, 'The Logic of Practical Discourse', in Contemporary
Philosophy, Vol. 1: Logic and Foundations of Mathematics (ed. by R. Kllbansky),
La Nuova Italia Editrice, Firenze, 1968, pp. 141-167.
[54] von Wright, Georg Henrik, 'On the Logic and Ontology of Norms', in Philosoph-
ical Logic (ed. by I. W. Davis, D. I. Hockney, and W. K. Wilson), D. Reidel

Publ. Co., Dordrecht, 1969, pp. 89-107.
[55] Aqvist, Lennart, 'Interpretations of Deontic Logic', Mind 73 (1964) 246--253.
[56] Aqvist, Lennart, 'On Dawson-models for Deontic Logic', Logique et Analyse 7
(1964) 14-21.
[57] Aqvist, Lennart, 'Choice-Offering and Alternative-Presenting Disjunctive Com-
mands', Analysis 25 (1965) 182-184.
NOTES
• Sections I-V of this paper are written by Risto Hilpinen on the basis of lecture-notes
by Dagfinn Ffllllesdal; sections VI-IX are written independently by Risto HiJpinen.
This work has been supported by grants from the Finnish Cultural Foundation
(Suomen Kulttuurirahasto) and the Finnish National Research Council for Social
Sciences (Valtion yhteiskuntatieteellinen toimikunta).
1 Cf. Quine [38], p. 387.
2 This expression has been used by C. D. Broad in [6].
3 See Fisher [11], Kanger [25], sections 2.3 and 2.4 (pp. 44-50 in this volume), von
Wright [53], p. 154, Aqvist [55], p. 248, and [57], p. 183 n. 1.
4 Cf. Anderson [1] and Castaneda [7].
5 Our notation here differs from Mally's; our u corresponds to Mally's U, n to n,
w to V, and m to /\.
6 In [30], this rule is applied e.g. on p. 29.
7 This is not the only possible explanation. In [19] Iaakko Hintikka has pointed out
that if the material implication q ::::> r in the antecedent of (AI) is replaced by a deontic
implication O(q ::::> r), we get an acceptable principle of deontic logic. Cf. also the
formula (41) below.
8 These papers, as well as the paper by GrelIing mentioned below, are reviewed by
Frederic B. Fitch in The Journal 0/ Symbolic Logic 5 (1940) 39-42.
9 Von Wright's 1951 system of deontic logic is described briefly also in [47], chapter 5.
10 These principles are theorems in all normal systems of alethic modal logic. For
various systems of modal logic, see Feys [10] and Hughes and Cresswell [23].
11 Cf. Iaakko Hintikka [19], p. 16, [20], and pp. 59-104 in this volume.
12 Cf. Thomas Aquinas, De Veritate, Q 17, art. 4, and Summa Theologiae, IaIIae,
Q 19, art. 6, and IlIa, Q 64.
13 We owe this term to Mr. Bengt Hansson; see [16) (p. 122 in this volume).
14 Cf. von Wright [46], pp. 1-2.
15 Cf. Iaakko Hintikka [19], pp. 7-9.
16 See e.g. [52], p. 16.
17 If this interpretation is accepted, the expression 'to do p' is, of course, somewhat
deviant. In [52) G. H. von Wright has suggested that it can be replaced by 'to see to it
that p' (p. 16). For simplicity, we shall use also the former expression regardless of the
interpretation of sentential letters.
18 cr. Hintikka [20] and pp. 79-82 in this volume.
1~ This weak notion of permission may also be termed 'permissibility' (cf. Hintikka
[20]). It must not be confused with the notion of right or claim as defined by Hohfeld
in [22], pp. 36--38. If a person a has against another person b a right (or a claim) that p,
then b has toward a a duty that p. Such obligations are not involved in the present
notion of permission. For a discussion of Hohfeld's analysis of basic legal notions, see
Kanger's paper in this volume, and Kanger and Kanger [26]. Another formalization
of Hohfeld's system can be obtained by relativising deontic operators to persons in
the same way as alethic modal operators are relativised in [IS]. If a's duty toward b
that p is expressed by '0 ... ~p', a's right (or claim) against b that p can be expressed by
'O~ ...P'. In the systems of deontic logic discussed in the present paper, deonticmodalities
are not relativised to persons. This involves the assumption that we are speaking only
of the obligations and permissions of one single person (or a group of persons).
20 (F2)-(F4) and (FS) correspond to the following condition of consistency: If a set
of sentences A is consistent and {O/I, Ols, ... , 0/,., Pg} £ A, then {Oil, Dis, ... , 01.. ,
/1, Is, ... , I,., g} is consistent.
21 The theory of deontic modalities presented by Hintikka in [20] corresponds to that
defined by the conditions (F1)-{F4), (FS), and (F9).
22 For the system M, see [10], p. 124.
23 There has been a great deal of discussion of the proper interpretation of the operator
Nin (GI). Cf. Anderson [3], pp. 348-354, [4], p. 111, and von Wright [53], p. 147.
24 In Levi's theory of induction and rational belief, 'the strongest sentence accepted
via induction' corresponds to Kanger's constant Q (cf. [29], pp. 29-30, and [17],
especially p. 84). In Hilpincn [17], the sentence h* corresponds to Q. This analogy
results from the similarity of deontic logic and the logic of rational belief (or 'accept-
ability'); the logic of rational acceptance sketched in Section 4.1. of [17] is similar to
the standard system of deontic logic.
25 In [9] E. E. Dawson has presented a reduction in which Olis defined as MNland
PI as NM/, and in [56] Lennart Aqvist has discussed other reduction schemata of the
same type. These 'reductions' do not appear very interesting from a philosophical
point of view.
26 For a discussion related to this point, see e.g. Grice [14], Section m.
27 cr. Anderson [4], pp. l08-U)9. Similar confusions seem to underlie the recent
discussion of 'choice-oft'ering' and 'alternative-presenting' disjunctive commands by
Rescher and Robinson [43], Gombay [12], and Aqvist [57]. For a criticism of this
discussion, see Bar-Hillel [5].
28 Cf. Prior [35], and [37], pp. 224-225; Chisholm [S], and von Wright [49].
29 Cf. Hansson [16] (pp. 132-133 in this volume).
30 cr. Hansson [16].
31 In [41] and [42] Nicholas Reacher has presented another system of conditional
deontic modalities. This system is discussed and criticized in detail by Bengt Hansson
[16] (reprinted in this volume); we shall not discuss it here.
32 Dr Audrey McKinney has pointed out in correspondence (in 1973) that according to the
truth-dt:finition (LJ) given below, 'O(rjr)' is always true, and hence
O(p& -pjp& -p)
is a logical truth. Thus (K2) should be weakened into
(K2*) 0 r => -(O(pjr)& O( - pjr»,
where 'Or' means that r is true in at least one possible world.
33 This bibliography includes only those works mentioned in this paper. A comprehensive
bibliography of deontic logic is to be found e.g. in von Wright [52].
STIG KANGER
NEW FOUNDATIONS FOR ETHICAL THEORY·
INTRODUCTION
There are philosophers who believe they really tell us how we ought
to behave, and there are philosophers who believe they ought to tell us
how we really behave. There are also philosophers of a more convenient
sort, who only wish to tell us something about ethical theories. The
doctrines advocated by these three kinds of philosophers all belong to
the wide domain of ethics, and they are easily confused. However, we
shall try not to confuse them here, and for that purpose, we may put
them under three distinct headings:
1. ethical theory proper
2. the psychology and sociology of morals
3. moral philosophy.
An ethical theory (of the proper kind) sets forth a system of ethical
propositions as true. Some ethical theories also set forth a system of
imperatives as correct. These propositions and imperatives are supposed
to give us the moral norms we are assumed to need.
We are all acquainted with an example of an ethical theory, namely,
the ten commandments.
An ethical theory may be more or less stringent. The stringency may
vary in several respects. For instance:
1. The propositions (and imperatives) set forth in the theory are all
formulated in a language with a certain formal structure. This structure
may be more or less specified.
2. The statements (and the imperative sentences) of this language are
either given or assumed to have a certain interpretation. This interpre-
tation may be more or less specified, and it may grant the propositions
(and the imperatives) a greater or less degree of lucidity.
3. Some statements (and imperative sentences) in the language are logi-
cally valid, and some are logical consequences of others. The methods for
demonstrating validity and consequence may be more or less developed.
R. Hllpinen (ed.). Deontic Logic: Introductory and Systematic Readings. 36-58. A.1I right. reserved.
Copyright © 1970 by D. Reidel Publishing Company. Dordrecht·Holland.
NEW FOUNDATIONS FOR ETHICAL THEORY 37
4. Some propositions (and imperatives) set forth in the ethical theory

are analytic. The set of these propositions may be more or less specified.
s. The set of propositions (and imperatives) set forth in the ethical
theory may be more or less specified and the methods for justifying them
may be more or less developed.
An ethical theory T 1 may be stronger or weaker than an ethical theory
T 2. If, for instance, each proposition (and imperative) set forth in T 2 is
entailed by the propositions (and imperatives) set forth in TlJ then Tl is
not weaker than T 2 •
We say that an ethical theory is true if all the propositions set forth
in the theory are true, and all the imperatives are correct.
Clearly, if there is a need for ethical theories at all, there is a need for
true ethical theories which are stringent and strong to a satisfactory
degree. In the attempts to obtain such theories, some of the raw material
may be supplied by the philosophers of our first two kinds. This raw
material consists of suggestions as to what propositions and imperatives
shall be set forth in the ethical theory. These suggestions do not, however,
form a satisfactory theory. The steps toward such a theory involve many
problems of logical, seman tical, and epistemological nature. It is the task
of moral philosophy to solve these problems.
Now and then, during the last few years, I have, with a steadily de-
creasing hope for success, tried to make a contribution to moral philos-
ophy so conceived. This paper is an abstract of some of my findings which
I publish before my hope is extinguished.
I shall assume that the reader has some familiarity with logic.
SEMANTICS AND THE PURE THEORY OF NORMS
1. A Formal Language
In this section, I shall outline a formal language fit for the part of
ethical theory which we call the pure theory of norms. This language,
which we call L, is obtained from the formal language of lower predicate
logic by the addition of a modal operator 'Ought' and an imperative
operator 'I'.
The symbols of L are the following:
(1) Parentheses and commas
(2) Propositional constants (to be introduced when needed)
38 STIG KANGER
(3) Name symbols:

Variables: Xl> X2' X3' •••
Constants: (to be introduced when needed)
(4) A symbol for identity: =
(5) Predicate constants with n argument places (n= 1,2, ... ) (to be
introduced when needed)
(6) Sentential connectives:
:::> (read: only if or if - then)
& (and)
v (or)
== (if and only if or if - then and only then)
,..., (not)
(7) Quantifiers:
U (for each)
E (for some)
(8) Modal operators:
Ought (it ought to be the case that)
(9) An imperative operator:
! (let it be the case that)
By an atomic formula of L, we understand either a propositional
constant or an expression having the form
(a = b)
or
/(al' a 2, ... , an)
where a, b, aI' ... , an (n= 1,2, ... ) are name symbols, and f is an n-place
predicate constant.
We define the class of formulas of L as the smallest class cP such that
(1) every atomic formula is in CP,
(2) if F and G are in CP, and if x is a name variable, then (F:::> G),
(F&G), (Fv G), (F== G), ,..,F, UxF, and ExFare in CP,
(3) if F is in fP and contains no occurrence of the imperative operator,

and if M is a modal operator, then MF and !F are in~.
The occurrences of name variables in a formula F will be classified
either as bound or as free in F. Thus an occurrence in F of x is bound in
F if it is contained in a formula UxG or ExG occurring in F, otherwise,
it is free in F.
We shall use the following definitions: A formula that contains no free
variable-occurrence is a sentence. If a sentence contains no occurrence
of the imperative operator, it is a statement, otherwise it is an imperative
sentence. A statement with no occurrence of modal operators is an
ordinary statement. A deontic statement is a statement which (in its
unabbreviated form) contains some occurrence of ·Ought'; and a
normative sentence is a sentence with some occurrence of 'Ought' or of
the imperative operator.
Finally, we shall introduce into L two new modal operators 'Right'
and ·Wrong' as abbreviatio'ns:
RightF for ,.., Ought,.., F
WrongF for Ought,.., F
·RightA' may be read as: 'it is right that A', and ·WrongA' as: ·it is
wrong that A'.
Deontic logic, i.e. a logic with some means for expressing the notion of
ought, originates with E. Mally, Grundgesetze des Sollens (Graz, 1926).
In this work Mally added a modal operator, denoting ought, to the
formal language of the propositional logic and laid down some (not very
happily chosen) axioms for this operator. Mally used the term UDeontik"
for what we call deontic logic.
The modalities Ought, Right and Wrong may be distinguished from
the predicates of being Obligatory, Permitted and Forbidden: the latter
apply to actions. A logical system involving these predicates has been
given in G. H. von Wright, 'Deontic Logic' (Mind 60 (1951) 1-15). The
statements of this system are built up by means of symbols for truth
functions and statements of the form PX expressing that some action of
the kind X is permitted (in short: X is permitted). The class of kinds of
actions is closed under the Boolean operations. Thus ',.., P X' and
, ,.., P ,.., X' may be interpreted as expressions of the facts that X is for-
40 STIG KANGER
bidden and X is obligatory. Ordinary statements are not "idmitted in von

Wright's system.
The imperative operator was introduced into logic in A. Hofstadter
and J. C. C. McKinsey, 'On the Logic of Imperatives' (Philosophy of
Science 6 (1939) 446-457).
2. Some Explanations and a Glimpse Beyond

In this section, I shall give a brief explanation of the notion of ought
which I have in mind, and which is denoted by the modal operator 'Ought'
in the language L. This explanation belongs to ethical theory rather than
to moral philosophy, and it is not an essential part of this paper. But
it may serve as a background which might make some of the subsequent
philosophical constructions more meaningful.
I shall also briefly explain some important ethical notions which are
easily confused with Ought, Right or Wrong, but which, nevertheless,
are of a different character. These notions lie beyond the scope of the
language L, but a preliminary explanation may perhaps prevent some
confusion.
Ordinarily we use the notations 'A', 'B', etc. as syntactic variables
denoting statements. In this particular section, however, we shall use
them as notations denoting propositions, i.e. statements paired with an
interpretation of the constants.
Consider the universe of discourse, and suppose there are welfare
programs for this universe. A welfare program consists of a set of
propositions expressing what is desirable for this universe from the view-
point of human welfare. We say that a welfare program is complete if
nothing that is relevant in connection with welfare has been overlooked.
We say that a welfare program is non-utopical if everything that it
proposes i.s possible to realize. We shall also distinguish what may be
called true welfare programs from alleged welfare programs. Now, the
proposition OughtA is true in the universe of discourse if and only if A
is entailed by each non-utopical, complete and true welfare program for
this universe.
It follows from this vague explanation that OughtA always implies
that'" A is avoidable. We note that the notion of avoidability is here
taken in a wide sense: only such facts are unavoidable which would be
or would have been outside the range of reasonable and foreseeingly
planned joint human efforts. Thus it might well be the case, for instance,
that it Ought to be so that Mr. X saves the drowning man in front of
him, even if Mr. X is unable to do so because he does not know how to
swim.
My explanation of Ought expresses a kind of utilitarianism (which
roughly corresponds to my own vague ethical beliefs). But we shall not
emphasize this point too much. We could very well interpret welfare in
such a way that the welfare programs would satisfy, for example, the
extreme deontologist, and our explanation would still serve its purpose.
So much for Ought. We have defined the modality Right as '" Ought",.
Thus defined it is related in a certain way to Ought. But, if we conceive
Right as an explication of the vague notion of rightness, this relation is
by no means a matter of course. There are other relations which seem to
be equally natural. We may say, to give a single example, that something
is right if it is not Wrong, and if it involves something that Ought to be
the case but that is not triVially obtained. To be more explicit: We say
that the proposition It is right that A is true in the universe of discourse
if and only if (i) the proposition RightA is true in this universe and
(ii) there is a proposition P such that the three propositions: it is un-
avoidable that if A then P, OughtP, and it is avoidable that P are true
in the universe of discourse. We note that rightness in this sense - we
may denote it by 'Right+' - admits indifferent facts that are neither
Right+ nor Wrong.
The difference between Ought and Right on one hand and Right + on
the other may be illustrated by an example. Consider the following three
propositions:
Al : Thomas sends five pounds to Elisabeth, if he owes five
pounds to Elisabeth.
A2 : Thomas sends a five-pound note to Elisabeth, if he owes five
pounds to Elisabeth.
A3: Thomas sends a five-pound note to Elisabeth without owing
her anything.
The three propositions OUghtAl and Right+ Al and RightA l are true
(we may argue). Right+ A2 and RightA 2 are true, but OUghtA2 is false,
since Thomas could equally well have sent five pound notes instead.
RightA3 is true but Right+ A3 and OughtA3 are false.
42 STIG KANGER
We shall not confuse the notion It is right that with the notion of a
Right which we may have in mind when making propositions like:
Elisabeth has a right to get back the five pounds she loaned to Thomas.
Propositions of this kind may be regarded as idiomatic instances of the
schema
(0) X has a right in relation to Y to the effect that F (X, V).
Here X and Yare what we might call 'moral personalities' (to coin an
analogy to 'legal personalities'). If we consider various instances of this
schema we shall find that the prefix 'X has a right in relation to Y to the
effect that' is ambiguous. The following four idiomatic instances of (0)
suffice to illustrate this fact:
(1) X has a right to get back the money she loaned to Y.
(2) X has a right to walk into V's shop when it is open.
(3) X has a right to give all her money to Y.
(4) X has a right to walk on the street outside V's shop.
In (1) 'right' means claim, in (2) 'right' means liberty or privilege, in (3)
'right' means power and in (4) 'right' means immunity. Corresponding to
these four senses of 'right' are four alternative explications of the ambig-
uous schema (0):
(1') Ought (Y sees to it that F(X, V»~
(2') Right "'" (X sees to it that ,.., F (X, V»~
(3') Right (X see~ to it that F (X, V»~
(4') Ought"'" (Y sees to it that"'" F(X, V»~
Finally, we shall not confuse Ought and Wrong with the notions of
praiseworthiness and blameworthiness. We usually apply these latter
notions to people (or to moral personalities). We may say, for instance,
that I am blameworthy for writing a paper on moral philosophy. This
means, I think, at least six things: (i) It is Wrong that I write a paper on
moral philosophy, (ii) I am writing a paper on moral philosophy, (iii) I
can avoid writing a paper on moral philosophy, (iv) I can know that it
is Wrong that I write a paper on moral philosophy, (v) I can know that
I am writing a paper on moral philosophy and (vi) I can know that I can
avoid writing a paper on moral philosophy.
We readily see that the notion of Right +, and the different senses of
a Right, as well as the notions of praiseworthiness and blameworthiness
NEW FOUND A TIONS FOR ETHICAL THEORY 43
cannot be analysed unless such concepts as: It is avoidable for X that,

X sees to it that and X can know that, are available. And since such
concepts are not available in the language L, we postpone all further
troubles to a planned second part of this paper.
In ethical literature the notions of ought, right and wrong are usually
conceived as predicates applying to actions - sometimes they are defined
only for actions which are free in some sense. A wrong action is usually
conceived as an action that ought not to be performed. But there is less
agreement as to the relation between right and ought. Two examples
- one on the Right line and one on the Right + line - might illustrate this.
G. E. Moore defines ought, right and wrong only for actions that are
voluntary in the sense that they could have been avoided, if we had chosen
to do so. A wrong voluntary action is one that ought not to be performed,
and a right voluntary actio~ is one that is not wrong. We note that there
are no indifferent voluntary actions. See G. E. Moore, Ethics (London
1912), Chapter I; cf. also Principia Ethica (Cambridge 1903), p. 148.
W. D. Ross suggests that a wlong action is one that ought not to be
performed, and a right action is one that is neither wrong nor indifferent,
in the sense that it does not fulfill any moral claim. (See W. D. Ross,
Foundations of Ethics, Oxford 1939, Chapter III.)
The distinction between claim, immunity, power and liberty, as differ-
ent senses of a right, is well known in jurisprudence. The distinction has
been particularly elaborated by W. N. Hohfeld. (See W. N. Hohfeld,
Fundamental Legal Conceptions, New Haven 1919, Chapters I-II.) The
Hohfeldian scheme may be carried over to the domain of ethics and it
may be outlined in a table with four groups, each containing four
equivalent schemata. (Instead of Hohfeld's term 'no-right' we use
'exposure', following a suggestion made in J. R. Commons, Legal Foun-
dations of Capitalism (Madison 1924). We also use the term 'claim' instead
of 'right'.)
(1) X has a claim in relation to Y to the effect that F (X, Y)

Y has a duty in relation to X to the effect that F (X, Y)
X has no exposure in relation to Y to the effect that,.., F (X, Y)
Y has no privilege in relation to X to the effect that,.., F (X, Y)
(2) X has a privilege in relation to Y to the effect that F (X, Y)
44 STIG KANGER
Y has an exposure in relation to X to the effr.ct that F (X, Y)

X has no duty in relation to Y to the effect that '" F (X, Y)
Y has no claim in relation to X to the effect that '" F (X, Y)
(3) X has a power in relation to Y to the effect that F (X, Y)
Y has a liability in relation to X to the effect that F (X, Y)
X has no disability in relation to Y to the effect that", F (X, Y)
Y has no immunity in relation to X to the effect that'" F (X, Y)
(4) X has an immunity in relation to Y to the effect that F (X, Y)
Y has a disability in relation to X to the effect that F (X, Y)
X has no liability in relation to Y to the effect that '" F (X, Y)
Y has no power in relation to X to the effect that '" F (X, Y)
I suggest that the schemata of (1)-(4) are equivalent to the schemata

(1 ')-(4') respectively. (See above.)
As to the distinction of ought and wrong from praiseworthiness and
blameworthiness - well, Aristotle, Nicomachean Ethics, bk. III: Chap-
ter 1, would head the reference list.
3. Valuation
I now turn to semantics. By a range, we shall understand a non-empty
class of individuals. We shall use the variable 'r' to denote ranges.
By a primary valuation (for L), we understand any binary operation V
(with the class of ranges as the first argument domain and the class of
constants and variables as the second) which, given a range r, assigns:
(1) 1 or 0 to each propositional constant,

(2) a member of r to each name symbol,
(3) a class of ordered n-tuples of members of r to each n-place
predicate constant. (n= 1, 2, ... We identify the ordered
I-tuple with its sole member.)
Thus, if P is a propositional constant, a a name symbol andfa 2-place

predicate constant, then VCr, P) is either 1 or 0, VCr, a) is a member of
r, and V( r,f) is a class of ordered pairs of members of r.
By the secondary valuation (for L), we shall understand a certain ternary
operation T, with the class of ranges as the first argument domain, the
class of primary valuations as the second, and the class of formulas as
the third. When r and V are given, T assigns either 1 or 0 to each formula
of L. The definition of T is recursive on the length of these formulas,

and it corresponds to the standard interpretation of the logical symbols:
(1) T(r, V,P)= V(r,P),
(2) T(r,V,(a=b»)=l ifandonlyif V(r,a)=V(r,b),
(3) T(r, V,J(ato ... , an») = 1 if and only if the n-tuple
(V(r, a1 ), ••. , VCr, an» is a member of V(r,J),
(4) T(r, V, (F => G») = 0 if and only if T(r, V, F) = 1 and
T(r, V, G) = O.
(5) T(r, V, (F & G») = 1 if and only if T(r, V, F) = 1 and
T(r, V, G)= 1,
(6) T(r, V, (F v G») = 0 if and only if T(r, V, F) = 0 and
T(r, V, G) = 0,
(7) T(r, V,(F == G») = 1 if and only if T(r, V, F) = T(r, V, G),
(8) T(r, V, '" F) = 1 if and only if T(r, V, F) = 0,
(9) T(r, V, UxF) == 1 if and only if T(r, V', F) = 1 for each
V' that is like V with the possible exception at x and r,
(10) T(r, V, ExF) = 0 if and only if T(r, V', F) = 0 for each
V' that is like V with the possible exception at x and r,
(11) T(r, V, MF) = 1 if and only if T(r', V, F) = 1 for each r'
such that RM(r', r),
(12) T(r, V, !F) = T(r, V,OughtF).
Explanation. We say that V' is like V with the possible exception at x
and r if (i) V' (r', P) = VCr', P) for each range r' and each propositional
constantP, (ii) V' (r',J) = V(r',J) for each r' and each predicate constant
f, (iii) V' (r', a) = V( r', a) for each r' and ef.zh name symbol a other than
x, and (iv) V'(r', x)= VCr', x) for each range r' other than r. The relation
RM will be explained in Section 6. M is a modal operator like Ought.
We shall use the following definitions: A range paired with a primary
valuation is a system. A model of a sentence S is a system (r, V) such
that T(r, V, S)= 1. The extension of S is the class of models of S. A
sentence S holds in a system (r, V) if T (r, V, S) = 1. If a statement A
holds in (r, V), then.A. is true in (r, V); A is/alae in (r, V) if A is not
true in (r, V). An imperative sentence C is correct in (r, V) if C holds
in (r, V); C is incorrect in (r, V) if C is not correct in (r, V). VCr, c)
is the denotation in (r, V) of the name constant c. Similarly, V(r,J) is
the denotation in (r, V) of the predicate constant f. A judgement is a
46 STIG KANGER
sentence paired with a primary valuation. A propositior is a statement

paired with a primary valuation. A proposition (A, V) is an ordinary
proposition if A is an ordinary statement; (A, V) is a deontic proposition
if A is a deontic statement. An imperative is an imperative sentence paired
with a primary valuation. A norm is a deontic proposition or an imper-
ative. A model of a judgement (S, V) is a range r such that T (r, V, S) = 1.
The extension of (S, V) is the class of models of (S, V). A jUdgement
(S, V) holds in a range r if T(r, V, S)= 1. A proposition (A, V) is true
in r if (A, V) holds in r; otherwise, (A, V) is false in r. An imperative
(C, V) is correct in r if (C, V) holds in r; otherwise, (C, V) is incorrect
in r. A name is a name constant paired with a primary valuation, and a
predicate is a predicate constant paired with a primary valuation. Ve"~ c)
is the denotation in r of the name (c, V), and VCr,!) is the denotation
in r of the predicate (f, V).
Valuation as a basic device of semantics originates with A. Tarski, 'Der

Wahrheitsbegriff in den formalisierten Sprachen' (Studia Philosophica 1
(1936) 261-405). A valuation in Tarski's sense involves analogies to what
we call systems and secondary valuation. The analogies to the systems
are sometimes called 'possible realizations'. A possible realization is an
ordered pair (r, v,.), where r is a non-empty class of individuals, and v,.
is a function (with the class of constants as the domain of arguments)
which assigns (i) 1 or 0 to each propositional constant, (ii) a member of
r to each name symbol and (iii) a class of ordered n-tuples of members
of r to each n-place predicate constant. The analogy to the secondary
valuation is a binary operation S, with the class of possible realizations
as the first argument domain, and the class of ordinary formulas as the
second. When a possible realizat,on is given, S assigns 1 or 0 to each
formula. Thus, for instance:
(1) S«r, v,.), P) = V,(P),

(2) S«r, v,.), (a=b»= 1 if and only if v,. (a) = v,.(b),
(3) S(r, v,.),!(a1 , ... , all» = 1 ifandonlyif
(v,. (a1), ... , v,.(all» is a member of Y,(f),
(4) S«r,v,.),-F)=l ifandonlyif S«r,V,),F)=O,
(5) S(r, v,.), (F::> G» = 0 if and only if S«r, v,.), F) =1
and S«r,v,.),G)=O,
(6) S«r, v,.), UxF) = 1 if and only if S«r, V;), F) = 1 for

each possible realization (r, V;) such that
(i) V;(P) = v,.(P) for each P, (ii) V;(f) = v,(f) for each 1 and
(iii) V; (a) = v,.(a) for each a other than x.
We say that an ordinary statement A is true, or satisfied, in (r, v,.) if
S«r, v,.), A)=1. We say that v,.(c) and v,.(f) are the denotations with
respect to v,. in r of c and 1 respectively.
Tarski's method of valuation does not seem to provide any simple
means for a suitable valuation of modal statements.
We may regard the valuation clauses for Ux and Ex as an explication
of the intuitive ideas of For each and For some. But then, perhaps, the
clauses are not quite evident. Instead of requiring that V' shall be like
V with the possible exception at x and r, we could, as it seems, make the
weaker requirement that V' and V must satisfy only (i)-(iii) of the four
conditions of the explanation above. Perhaps we could require something
still weaker: V' and V must satisfy (i)-(iii) only when r' =r, and not
necessarily for each r'. These changes make no difference as long as only
ordinary formulas are in question. But when modalities are involved, we
are faced with the problem of choosing between our original valuation
of quantification and the two revised versions. (Cf. S. Kanger, 'A Note
on Quantification and Modalities', Theoria 13 (1957) 133-134.) Now, I
think there is an argument against a revision. Geach has noted that
Ex RightFx (where Fx is an ordinary formula) is intuitively sp-onger than
RightExFx. (See A. N. Prior, Time and Modality, Oxford 1957, p. 142.)
Hence, by a correct valuation, it ought to be so that
T(r, V, (Ex RightFx => Right ExFx» = 1
always holds but
T(r, V, (Right ExFx => Ex Right Fx» = 1
will not always hold. And this will be the case only if we retain our
original valuation.
The valuation of ! F and the notion of correctness are roughly suggested
in A. Hofstadter and J. C. C. McKinsey, 'On the Logic of Jmperatives'
(Philosophy 01 Science 6 (1939) 446-457). (Hofstadter and McKinsey also
suggest and elaborate another interpretation of ! F, which justifies the
valuation: T(r, V, !F)=T(r, V, F).)
48 STIG KANGER
The notion of proposition as we have defined it is akin to the notion

of proposition in traditional logic. This notion may perhaps be explicated
as a couple (A, VA)' where VA is a primary valuation confined to the
symbols actually occurring in A. An indication of this notion is found in
Aristotle, De Interpretatione, Chapter 4. It underwent refinement in
mediaeval logic (cf. A. Church, 'Propositions and Sentences' in The
Problem of Universals (ed. by I. M. Bochenski et al.), Notre Dame 1956).
A proposition in our sense should not be confused with the intensional
entity constituting the meaning or sense of a proposition. Such entities
originated with the Stoic-Megaric logic (cf. B. Mates, Stoic Logic,
Berkeley and Los Angeles 1953), and they were given a revival in modern
semantics in B. Bolzano, Wissenschaftslehre (1837), and in G. Frege,
'Ober Sinn und Bedeutung' (1892). We may, perhaps, identify the sense
of a proposition with the extension of the proposition.
4. Some Further Notions of Semantics

We shall use the following definitions: A sentence S is valid if S holds in
every system. S is contravalid if S holds in no system. S is contingent if
S is neither valid nor contravalid. S is a logical consequence of a (possibly
empty) sequence r of sentences if S holds in every system in which all
the members of r hold simultaneously. Sand r are logically equivalent
if S holds in exactly those systems in which the members of r hold
simultaneously. A judgement (S, V) is analytic if (S, V) holds in every
range. (S, V) is contradictory if (S, V) holds in no range. (S, V) is
synthetic if (S, V) is neither analytic nor contradictory. (S, V) is
entailed by a (possibly empty) sequence L1 of judgements if (S, V) holds
in every range in which all the members of L1 hold simultaneously. (S, V)
and L1 are synonymous if (S, V) holds in exactly those ranges in which
the members of L1 hold simultaneously. Two names (c, V) and (c', V')
are synonymous if V(r, c)= V'(r, c') for each r. Similarly, two predicates
(f, V) and (f', V') are synonymous jf V(r,f)= V' (r,f') for each r.
The definitions of validity and logical consequence derive their essential

features from A. Tarski, 'Ober den Begriff der logischen Folgerung'
(Actes du Congres International de Philosophie Scientijique, Vol. 7, Paris
1936, pp. 1-11). (We may note that Tarski uses the term 'analytic' in the
sense of validity.) Tarski defines a valid ordinary statement as an ordinary
statement that is satisfied in every possible realization. Clearly, an ordi-

nary statement is valid by Tarski's definition if and only if it is valid by
ours. Tarski's definition of logical consequence is related to ours in the
same way as his definition of validity is related to ours.
The valuation of ! A has been chosen in such a way that! A and OughtA
are always logically equivalent. This fact is, I think, in accordance with
our preconceptions of the matter. On the other hand, the alternative
assumption that! A and A are always logically equivalent seems to violate
these preconceptions. Hence the valuation T(r, V, !F)=T(r, V, F) for
! F and consequently the ideas underlying the imperative logic of Hof-
stadter and McKinsey seem to be inadequate.
In any event, an acceptance of the Hofstadter-McKinsey kind of inter-
pretation of imperatives is awkward if it is combined with a confusion of
imperatives and deontic propositions. A good illustration of this may be
found in G. H. von Wright, 'Om s.k. praktiska slutledningar' (Tidsskrift
for Rettsvitenskap 68 (1955) 465-495). According to von Wright, a norm
such as X is obligatory or It ought to be so that X is performed (where
X is an act or a class of actions) is satisfied if X is performed; otherwise
it is dissatisfied. A norm such as X is permitted is satisfied if and only if
X is performed. The norm X is permitted cannot be dissatisfied, but it
may be not satisfied. A norm N1 is said to follow from the norm N z and
a statement A if (i) the satisfaction of N1 follows from the satisfaction of
N z and the truth of A and (ii) the dissatisfaction of N z or the falsity of A
follows from the dissatisfaction of Nt> provided N1 can be dissatisfied.
By these definitions, we conclude that the norm X is obligatory follows
from the statement X is performed and the norm X is permitted.
Our definition of analyticity may be regarded as an explication (and
an extension to imperatives) of the idea that an analytic proposition is a
proposition that is true in every possible universe. In philosophical
literature, we may find at least two other notions (or main types of
notions) of analyticity. The first is due to Kant and the second to Bolzano.
The following is a sophisticated but, I think, fairly adequate explication
of Kant's notion: A proposition (A, V) is analytic if and only if it is
analytic (in our sense), and V is a standard primary valuation for L (in
the sense to be explained in the next section). Cf. E. W. Beth, 'Kant's
Einteilung der Urteile in analytische und synthetische' (Algemeen Neder-
lands Tijdschrift voor Wijsbegeerte en Psychologie 46 (1953-54) 253-264).
50 STIG KANGER
The second notion of analyticity may be indicated thus: By a logically

true statement we understand a statement A such that the result of gener-
alizing all extralogical constants in A is true. By an analytic statement we
understand a statement which is synonymous with or equivalent by
definitions to a logically true statement. The essence of this idea must be
credited to Bolzano. See B. Bolzano, Wissenschaftslehre II (1837), § 148.
Among later adherents we may particularly note G. Frege and W. v.
Quine. See G. Frege, Grundlagen der Arithmetik (Breslau 1884), pp. 3f.
and W. V. Quine, 'The Problem of Interpreting Modal Logic' (The
Journal of Symbolic Logic 12 (1947) 42-48). Frege gives this version of
the idea: A statement is analytic if and only if it is deducible by means of
(higher) logic from a system of definition statements. If we adopt Kant's
notion, the so-called real definitions are always analytic, but nominal
definitions and principles of arithmetic are not. If, on the other hand, we
adopt the notion of Bolzano and Frege, all definition statements (or in
any event, all definition statements belonging to the system of definitions
subjoined to logic) and all principles of arithmetic are analytic.
5. The Language (L, cp >

There are several usages of the formal language L. In each usage, we
assume a certain range of discourse and a certain complete interpretation
of the constants and the free variables. Thus, each usage corresponds to
a certain system. We regard some of the usages of L as the standard
usages. There may be several standard usages of L. Several assignments
of a denotation to a constant may, for instance, be consistent with
standard usage because of the ambiguity of the constant. We shall note
also that L's constants may be vague in the standard usage: we are not
necessarily required to know the denotations of the constants.
Now, let cp be the class of standard systems, i.e. the class of systems
which correspond to the standard usages of L.
By a language, we shall understand - roughly speaking - a formal
language paired with a class of systems which correspond to the standard
usages of the formal language. Thus (L, cp) is a language. We shall make
the assumption that every system in cp contains the same range - we may
call it the standard range (for I) and denote it by 'r*'. We may roughly
describe r* as the class of all real entities. We shall call the primary
valuations contained in the systems of cp the standard primary valuations
(for L). Let V* be a standard primary valuation. We shall assume now

that the class of standard primary valuations is the class of all V such
that V(r, c)= V*(r, c), V(r,j) = V*(r,j) and V(r, P)= V*(r, P) holds
for each range r and for each name constant c, predicate constant f and
propositional constant P. In other words, we shall assume that every
constant is unambiguous and that every variable is completely ambiguous
in (L, qJ).
In the sequel, let 'V*' be a variable for standard primary valuations.
We say that a name (c, V*) of (L, qJ) is vague in r if there is no
effective method available to decide for each member x of r whether or
not X= V*(r, c). We say that an n-place predicate (J. V*) is vague in r
if there is no effective method available to decide for each ordered n-tuple
(Xl' """' Xn) of members of r whether (xl> """' xn) is a member of V* (r,f).
We also say that a judgement (S, V*) is vague in r if there is·no effective
method available to decide whether T(r, V*, S)= l.
Clearly, one source of vagueness in (L, qJ) is an incomplete specification
of qJ. Another source is an incomplete specification of the relation RM
occurring in the valuation of modal statements. We shall not, however,
regard a complete specification of qJ and RM as a desideratum. On the
contrary, we shall think of qJ and RM as if they were specified in such a
way that the names, predicates and judgements of (L, qJ) acquire the
vagueness they have in the standard usages of L - so to speak. (There are
also other and deeper sources of vagueness, but we may leave these
without notice in this connection.)
Some further definitions: A sentence S holds, a statement A is true
(false), and an imperative sentence C is correct (incorrect) if S holds, A
is true (false) and C is correct (incorrect) in some standard system
(r*, V*). S is analytic if some judgement (S, V*) is analytic. A judge-
ment (S, V*) holds, a proposition (A, V*) is true (false), and an imper-
ative (C, V*) is correct (incorrect) if (S, V*) holds, (A, V*) is true
(false) and (C, V*) is correct (incorrect) in the standard range r*. We
say that X is the denotation of a name constant c and of name (c, V*)
if X= V* (r *, c). X is the denotation of a predicate constant f and of a
predicate (f, V*) ifX= V*(r*,f).
Note that these definitions of truth, denotation, etc. profit from the
unambiguity of the constants in (L, qJ) and from the uniqueness of the
standard range.
52 STIG KANGER
6. Normative Logic
Let the symbol 'M' denote anyone-place modality. Assume that
T(r, V, MA) is defined for each r and V and each statement A of L. Let
.£I be any empty or non-empty, finite, denumerable or non-denumerable
sequence of propositions of L. Let MLI be the sequence obtained from .£I by
prefixing 'M' to every statement contained in the proposition of .£I. Thus,
if .£I is (BI' VI), (B2' V2), then MLI is (MBl> VI), (MB2' V2).
Now the following two conditions are always equivalent:
(1) For each .£I and each (A, V), if .£I entails (A, V), then MLI
entails (MA, V).
(2) There is a relation R such that for each r and each (A, V),
T(r, V, MA)= 1 == (r') (R(r', r) ~ T(r', V, A)= 1).
The relation R is unique:
If for each r and (A, V)
T(r, V, MA) = 1 == (r') (R(r', r) ~ T(r', V, A) = 1),
then for each r' and r
R(r', r) == (B)(V)(T(r, V, MB) = 1 :::> T(r', V, B) = 1).
If M is Ought and if R is Roupt> then the first result provides a justifi-
cation of the valuation for the deontic statements. The second result
provides a kind of explanation of Roulh,. Perhaps we may paraphrase it
as follows: Roullht is the relation which holds between any two universes
r' and r such that every proposition that ought to be true in r is true in r'.
We shall now make three assumptions concerning the relation Roulht:
I (r)(Er') ROullhl(r', r)
II (Er) "'" ROup,(r, r)
III (Er')( (Er) ROulhl (r', r) & (Er) "'" ROupt (r', r»)
Assumption I is equivalent with the assumption that
(OughtA:::> RightA)
is always valid. Assumption III is equivalent with the assumption that
there are synthetic deontic propositions of the form (OughtA, V). It is
also clear that III is an expression of a kind of moral relativism.
NEW FOUNDATIONS FOR ETHICAL THEORY S3
I shall now list a few valid sentences. The validity of these sentences
does not depend on the assumptions just made.
1 Ought A, where A is valid

2 (Ought(A B) :::> (OughtA :::> Ought B»)
:::>
3 (Ought(A & B) == (OughtA & Ought B»

4 (Right(A :::> B) == (OughtA :::> Right B»
5 (Right(A v B) == (RightA v Right B»
6 (OughtUxFx :::> OughtF~)
7 (OughtUxFx :::> UxOughtFx)
8 (ExOughtFx :::> OughtExFx)
9 (RightUxFx ::J Ux RightFx)
10 (Ex RightFx :::> RightExFx)
11 (OughtA == ! A)
We note that the converses of 7 and 10 are not always valid. This fact
depends primarily on the properties of quantification. Cf. the note at the
end of section 3. We note also that
(a = b) :::> (OughtFa :::> OughtF:»
is not always valid. An assumption to the contrary would lead to a para-

dox of the same kind as the well-known Morning Star paradox.
We may, perhaps, deny the truth of assumption III. A denial of III is

equivalent with the assumption that 'Ought' is definable thus:
OughtA ==df N(Q :::> A)

where N is the notion of analytic necessity:
T(r, V, NB) = 1 == (r') (T(r', V, B) = 1)

and Q is a propositional constant with a fixed valuation:
T(r', V, Q) = 1 == (r) ROuBht(r', r)
(We may think of Q as a constant stating what morality prescribes.) If
S4 STIG KANGER
we accept this definition of 'Ought', we may obtain a deontic logic by

extending a logic for analytic necessity with Q as a new primitive symbol
and with the statement ('" NQ &'" N '" Q) as a new postulate. Such a
logic was given in a paper on deontic and imperative logic (including a
theory of unavoidability), which I wrote in 1950 and submitted in partial
fulfillment for the lic. phil. degree at the University of Stockholm. Almost
the same kind of deontic logic has been given in A. R. Anderson, The
Formal Analysis of Normative Systems (New Haven 1956). A summary
of some main ideas in Anderson's paper may be found in A. R. Anderson
and o. K. Moore, 'The Formal Analysis of Normative Concepts'
(American Sociological Review 22 (1957) 9-17) and in A. N. Prior, Time
and Modality (Oxford 1957), Appendix D. I am now inclined to reject
this definition of Ought; my main reason is the fact that some deontic
propositions (OughtA, V) seem to be synthetic. But I also think that
the vagueness of such deontic propositions excludes the possibility of
making a definite decision in this case.
There is another assumption which perhaps might be adopted:
(r)(r')(r")(ROuBhtCr', r) :::> (ROuRht(r", r') == RouJht(r", r)))
If we assume assumption I, this assumption is equivalent with the assump-

tion that
Ought A :::> Ought Ought A
and
Right A :::> Ought Right A
are always valid. It implies that
Ought(OughtA :::> A)
is always valid. Thus, it may be regarded as an expression of what might

be called moralism.
7. A Dialogue
The complete moral philosopher: Excuse me for interrupting you, Mr.
Kanger, but would you admit a short interview before you proceed?
Kanger: Yes.
Ph: According to a wellknown theory in moral philosophy, known as
the emotive theory, deontic propositions are neither true nor false. Now,
I understand, you have the opposite view.
K.: Yes.
Ph.: Of course you and the adherents of this theory may have in mind
two different notions of truth. But your notion of truth seems to be in
agreement with scientific semantics and I am sure that the adherents of
the emotive theory would adopt it if they were met with the problem.
K.: Yes.
Ph.: Now, clearly, deontic propositions with the valuation they got in
Section 3 must be either true or false. So, if the adherents of the emotive
theory wish to sustain their standpoint they have to reject the valuation
clause for Ought or the equivalent thesis that OughtLi always entails
(OughtA., V) when LI entails (A., V), which you gave in Section 6.
K.: Yes.
Ph.: The equivalent thesis can, of course, be refuted by the argument
that deontic propositions are neither true nor false.
K.: Yes.
Ph.: I realize that this would be a petitio principii. But perhaps there
are other arguments for rejecting the valuation clause or its equivalent.
Let me try the argument that deontic propositions of the form
(Ought A. , V) do not state anything about reality. But this argument
involves, of course, nothing that might refute the valuation clause or the
idea that deontic propositions are true or false. At most, we may conclude
that deontic propositions are not synthetic.
K.: Yes.
Ph.: The feeling we may have that (Ou~htA., V) does not state any-
thing about reality is perhaps easily explained: Because of the vagueness
of Ought we may feel that (OughtA., V) is not synthetic, even if it is so,
and hence, we may also feel that it does not state anything about reality.
K.: Yes.
Ph.: Then I know of no other argument against your standpoint except
the one that deontic propositions are synonymous with imperatives. But
I also know your answer: Every deontic proposition is synonymous with
an imperative, and this fact is in full agreement with everything else in
this paper.
K.: Yes.
Ph.: Now let me return to the notion of truth. Perhaps our agreement
S6 STIG KANGER
on this point was a little rash. Could we not restrict the application of the
truth predicate to, say, non-deontic statements and call the formerly true
deontic statements correct instead? And couldn't we do this and still be
in agreement with scientific semantics?
K.: Yes.
Ph.: Do you mean that the choice of the range of applicability of the
truth predicate is, to some degree at least, conventional?
K.: Yes.
Ph.: So the adherents of the emotive theory have a chance to be right
after all by a terminological trick.
K.: Yes.
Ph.: I now turn to a new problem. There is a .wellknown distinction
between so called natural properties and non-natural properties. Some
authorities believe that value is a non-natural property, while others
disagree. In this paper, the distinction is difficult to maintain because of
your tendency to do away with all kinds of spurious entities. The predi-
cates do not refer to properties in the sense we may have in mind in this
connection, but to classes of individuals or to classes of ordered sets of
individuals. And I cannot see how one class can be less natural than an-
other. This fact does not, of course, mean that the distinction between
naturalism aJ?d non-naturalism cannot be maintained at all. We may
perfectly well distinguish between naturalistic and non-naturalistic state-
ments and propositions. Thus, we say that A is naturalistic if A or ,.., A
is logically equivalent with a sequence of ordinary statements, and non-
naturalistic otherwise. We say that <A, V*) is naturalistic if <A, V*) or
<,.., A, V*) is synonymous with a sequence of ordinary propositions of
<L, (fJ), and non-naturalistic otherwise.
K.: Yes.
Ph.: It follows from assumptions I and II of the preceding section that
there are non-naturalistic deontic statements. But we may raise the
problem whether or not there are non-naturalistic deontic propositions
of <L, (fJ). The answer to this problem depends on the choice of (fJ.
However, I think we need not hesitate to assume that there are non-
naturalistic deontic propositions of <L, (fJ). In any event, if we turn to
standard English and raise the analogous problem, then a negative
answer would certainly be unwarranted - it would be, we may argue, an
expression of the so called naturalistic fallacy. But the naturalist position
is easier to defend on the meta-level, so to speak. For each deontic state-

ment A. and each (r, V) there is a naturalistic proposition in the meta-
language, which states a necessary and sufficient condition for the truth
of A. in (r, V).
K.: Yes.
Ph.: There is another problem which should not be confused with the
preceding one: Is there a statement OughtA. that is a logical consequence
of a contingent ordinary statement and that also has a contingent ordinary
statement as a logical consequence? An affirmative answer to this
question would mean that we can draw ethical conclusions from non-
ethical premisses in a non-trivial sense.
K.: Yes.
Ph.: According to assumption I of the preceding section, there are
contingent statements of the form OughtA.; but, by saying this, we do
not provicle a strict answer. to our questions. Perhaps we should leave it
undecided and allege the vagueness of 'Ought' as an excuse.
K.: Yes.
Ph.: I have a final problem. The thesis equivalent with the valuation
for 'Ought' and given in the preceding section seems to me to be indubi-
table. Hence we have to accept the valuation. Furthermore, (Ought A.
~ RightA.) is clearly always valid, and hence, we have to accept assump-
tion I. But this assumption implies, roughly speaking, that there is a
universe r' which is a moral standard for our universe r*.
K.: Yes.
Ph.: But what is this universe, if I may inquire? Heaven? Or do I have
to review my thoughts on this ultimate ma.tter once more?
K.: Yes.
Ph.: Well, good-day.
We must forgive our complete philosopher for his incomplete references.

The emotive theory originates with A. Hagerstrom, Om moraliska
fiirestallningars sanning (Stockholm 1911). A later, but independent,
expression of the theory is given in C. K. Ogden and I. A. Richards, The
Meaning of Meaning (London 1923), p. 125. Still later versions of the
theory may be found, for instance, in B. Russell, Religion and Science
(New York 1935), Chapter 9; A. J. Ayer, Language, Truth, and Logic
(London 1936); Chapter 6, I. Hedenius, Om ratt och moral (Stockholm
S8 STIG KANGER
1941); C. Stevenson, Ethics and Language (New Haven 1914); A. J. Ayer,

'On the Analysis of Moral Judgments' (Horizon 20 (1949»; and R. M.
Hare, The Language of Morals (Oxford 1952). The importance of the
emotive theory lies particularly in its emphasis on the phenomenon of
emotive meaning. Thus, the emotive theory cannot be properly charac-
terized as the theory which denies truth and falsehood to ethical judg-
ments. The phenomenon of emotive meaning cannot be analyzed unless
we consider the concrete instances of expressions in the context of human
communication, but that is not the concern of this paper.
The prominence of the naturalistic fallacy originates with G. E. Moore,
Principia Ethica (Cambridge 1903). The idea that we cannot draw ethical
conclusions from factual premisses goes back to David Hume, A Treatise
of Human Nature, Book 3 (1740), Part 1, Section 1.
The non-naturalist ethics and moral philosophy has been developed in
G. E. Moore, Principia Ethica (Cambridge 1903) and Ethics (London
1912), and in D. Ross, The Right and the Good (Oxford 1930) and Foun-
dations of Ethics (Oxford 1939) (to mention some of the most important
contributions only).
Among recent naturalistic works we may note R. B. Perry, General
Theory of Value (New York 1926) and B. Russell, Human Society in
Ethics and Politics (London 1954).
The naturalist-non-naturalist controversy and the controversy about
the emotive theory have given rise to many studies. We may mention
particularly G. E. Moore, 'A Reply to My Critics' (in The Philosophy of
G. E. Moore (ed. by P. A. Schilpp), Evanston and Chicago 1942, pp. 535-
677). Other studies may be found in W. Sellars and J. Hospers, Readings
in Ethical Theory (New York 1952).
University of Uppsa/a
NOTE
• An earlier version of this essay was published as a privately distributed pamphlet:
New Foundations for Ethical Theory. Part 1, Stockholm 1957.
JAAKKO HINTIKKA
SOME MAIN PROBLEMS OF DEONTIC LOGIC
I. THE STRUCTURE OF FUNDAMENTAL NORMATIVE CONCEPTS
1. In deontic logic, we must be able to discuss various predicates or

attributes (properties and relations) of (human) acts. As place-holders for
expressions denoting such predicates, we shall in this paper use the upper-
case letters A, B, C, ... , Q, R, .... The reader may think of them as
standing for some (unspecified) properties and relations of human acts.
It is important to realize, and to keep in mind, that these letters do not
represent individual acts, but stand for certain general characteristics of
such act-individuals (or for general characteristics of n-tuples of such
individual acts).
In addition to these place-holders we shall use the familiar proposi-
tional connectives '....., , (no), '&' (and), and 'v' (inclusive or). Other
propositional connectives, including' => ' (if-then) and' ==' (if and only if),
are assumed to have been defined in terms of the first three connectives
in the usual way. The fundamental deontic concepts will be expressed by
the operators '0' (for obligation or duty), 'F' (for forbiddance or pro-
hibition) and 'P' (for permission).
Approximately - but only approximately - these basic deontic oper-
ators could be read as follows:
Op = it ought to be the case that p

= it is obligatory that p
= it must be the case that p
Fp = it is forbidden that p
Pp = it is permitted that p
= it may be the case that p
The concept of permission is nevertheless so ambiguous that in many
contexts it is safer to adopt the following. more cautious reading:
Pp = it is permissible that p.
R. HliplM1I (ed.), Deo1ltlc Logic: I11t.odIu:to.y tIIId SystemAtic Readings, 59-104. All rights .ese,.ed.
Copy.ight C> 1970 by D. Rdtkl Pooblishi1lg COmp01ly, Dordrecht-Holltllld.
60 JAAKKO HINTIKKA
In this paper we shall disregard the (highly relevant) fact tltat obligations,
and even more so duties, are usually thought of as obliging some particular
person, and that permissions often are similarly 'personal'. In this chapter,
and in its immediate sequel, the reader will have to think of obligations
in completely 'impersonal' terms. Most of our discussion will nevertheless
remain valid if it is assumed that all the obligations, permissions, prohi-
bitions, etc. we are considering relate to one and the same person. Whether,
and if so how, 'personal' obligations and permissions can be defined in
terms of the 'impersonal' ones is an interesting question which will not
be discussed here.
As bound (bindable) individual variables we shall use the lower-case
letters x, y, .... As quantifiers we shall employ the usual symbols '(3 x)'
and '(y)' (where x and y may of course be replaced by other bindable
individual variables). As place-holders for expressions referring to
individual acts (as 'names of act-individuals') we shall use the lower-case
letters a, b, c, ....
2. It is implicit in what was just said that the individuals we shall in the
first place consider are individual acts. It would be easy to introduce
variables and quantifiers ranging over some other kinds of individuals,
for instance Qver persons. By using such variables and quantifiers, im-
portant further observations can be made and codified. It seems to us
that these observations are nevertheless not quite as fundamental as
those that can be made in terms of variables ranging over act-individuals
(and in terms of place-holders for expressions which refer to such in-
dividual acts).
But must we really - and may we - use variables of the latter kind, that
is to say, deal with particular acts as being (logically speaking) individuals
over which we can quantify? Here I cannot do full justice to these ques-
tions. Fortunately, it so happens that Donald Davidson has argued my
case for me to a considerable extent in his paper 'The Logical Form of
Action Sentences' (in The Logic of Decision and Action, ed. by N. Rescher,
University of Pittsburgh Press, Pittsburgh, 1967, pp. 81-95). There he
argues incisively for the affirmative answer both to the question of
possibility and to the question of necessity of considering particular acts
as (logical) individuals. Suffice it here to refer the reader to Davidson's
arguments. It may be added, though, that insofar as we can elucidate
SOME MAIN PROBLEMS OF DEONTIC LOGIC 61
deontic problems by considering acts as individuals, this success will

reinforce Davidson's point.
The letters p, q, ... (with or without a subscript) are metalogical symbols
which serve as place-holders for arbitrary formulae. The expression 'p (alb)'
will be used to stand for the formula obtained from p by replacing every
(free) occurrence of 'b' by a (free) occurrence of 'a'. The same notation
will be applied to bound variables.
By means of these conventions, we can formulate the rules of formation
for our system in the following way:
(F. I ) An n-adic predicate followed by a pair of parentheses enclosing
n free individual variables is a formula (formulae of this kind
are called atomic);
(F.2) If p and q are formulae, then so are '"" p, (p & q), (p v q),
Op, Pp, and Fp;
(F.3) If p is a formula containing the free individual variable 'a'
but not the bound variable 'x', then (3x)p(xla) and
(x) p(xla) are also formulae.
In the rule (F.3), the variables 'a' and 'x' may be replaced by other
variables of the same kind (free and bound, respectively). In the same way,
rules which are formulated in terms of particular variables are in the
sequel understood to apply to all variables of the same kind.
The rules (F.I}-{F.3) define the formulae we are going to use for our
purposes. In discussing the systems of other authors, however, we shall
frequently quote formulae of different kiIlds, too. For instance, in the
above rules the deontic operators '0', 'P', and 'F' are applied to formulae
and not directly to predicates of acts, while in other systems such a direct
application is allowed. The merits of our procedure are shown by the
arguments by means of which we shall try to show (in Sections 5 and 6
below) the indispensability of quantifiers in deontic logic.
3. As the reader may gather from the introductory survey paper of this
volume, most expositions of deontic logic fail to utilize quantifiers, i.e.
the precise logical counterparts of such words as 'every' and 'some'.
Sometimes quantifiers are brought into treatments of deontic logic as
additions to systems already constructed. Even when quantifiers are
62 JAAKKO HINTIKKA
treated as such appendages, only a couple of serious attf'mpts are made

to develop a serious semantical theory of quantified deontic logic.
It seems to us, nevertheless, that the use of quantifiers in deontic logic
has a much deeper significance than that of a way of extending already
existing systems. The introduction of quantifiers is in fact already
necessary for the purpose of analyzing adequately some of the funda-
mental concepts (and their interrelations) which obviously fall within the
purview of every system of deontic logic. Among them, there are the
concepts of obligation ('ought to', the 'normative must'), of prohibition,
of permission (the 'normative may'), and of commitment. It is our thesis
that in certain uses of these concepts which fall within the intended scope
of existing treatments of deontic logic, their logical interrelations cannot
be expressed without using the words 'every' and 'some' or some other
logical expressions in terms of which they are definable.
4. In the current systems, formulae of the following forms are provable:
(1) -0- p == Pp
(2) -P- p == Op
(3) -Op == P-p
(4) -Pp == O-p
(5) Op == F-p
(6) Fp == O""p
Roughly speaking, (1) may be read 'it is permissible thatp if and only
if it is not obligatory to omit p', (2) may be read 'p is obligatory if and
only if it is not permissible that not-p', and correspondingly for the other
types of formulae. In virtue of (1), 'P' may in the current systems be
replaced everywhere by''''' 0 -', and vice versa; and similarly for (2)-(6).
In virtue of (6), 'F' can always be eliminated in terms '0' and' - '.
I do not want to criticize (1)-(6) or the related possibilities of elimi-
nating '0', and 'F' and 'P' in terms of one another. In fact, I shall
assume the same interdependencies in my own approach. What I would
like to point out are the awkward consequences of replacing 'p' in (1)-(6)
by a variable ranging over properties of acts. In other words, I maintain
that one cannot subscribe to (1)-(6) while employing formulae like OA,
FA, and PA as the formal counterparts of the statements 'A is obligatory',

'A is forbidden', and 'A is permissible', where A is to be replaced by an
attribute of acts. In still other words, I would say that the logical structure
of obligations, prohibitions, and permissions pertaining to kinds of acts
(in contradistinction to individual acts) is more complicated than what
is usually assumed.
5. What, then, is this structure? In answering the question it is advisable

to start from the notion of prohibition. It is the clearest of all the funda-
mental deontic notions. Obviously, saying that acts of a certain kind A are
forbidden is to say that no act under consideration may be of this kind.
In other words, every act under consideration ought to be an omission
of A, i.e. every act ought to be an instance of the negation - A of A.
We may hence express that A is forbidden by a sentence of the form
(7) (x)O-A(x).
(Whether and when (7) can be strengthened to
(8) O(x)-A(x)
will be discussed later.)
This takes care of the notion of prohibition. How is it related to the
other notions? If (7) is taken to be an explicatum of FA and if the usual
relation between 'F' and '0' is taken to hold, then OA should be expli-
cated by (x) 0 - - A (x), which reduces to
(9) (x)OA(x).
In more intuitive terms this means that, according to the current assump-
tions, we are obligated to do A only if every act of ours ought to be an
instance of A (where 'A' may be replaced by any attribute of acts). There
are attributes of acts for which this is not unnatural. For instance, one
can say that 'you ought to abstain from stealing' means 'every act of
yours ought to be an instance of abstention from stealing'. However,
obligations of this kind are not the only ones. Perhaps they are not even
primary, for they seem to me to be derived from some prohibition.
Most of the ordinary 'positive' obligations exhibit a different kind of
logical behavior. When it is said that each year one ought to pay one's
income tax, this does not mean that each year one's every act ought to
64 JAAKKO HINTIKKA
be an instance of taxpaying. Rather, it means that each year the act of

paying the income tax ought to be among the things one does. Similarly,
the (moral) obligation of helping a needy neighbor may be fulfilled by
one act, presupposed that this act suffices to remove the need.
The logical form of obligations of this kind is
(10) O(3x)A(x)
which is different from (9). The logical relation of this notion of 'positive'
obligation to prohibition is not the one it is usually assumed to be. In
fact, if'OA' and 'FA' are taken to be shorthands for (10) and (7), respec-
tively, then '0' and 'F' cannot be defined in terms of each other simply
by means of '-'.
In addition to the analysis of obligations and prohibitions, we need
quantifiers in order to analyze the notion of permission. Again, the usual
relations between the basic deontic notions fail. Supposing that the
notion of permission is related to the 'negative' obligation (9) in the way
it is usually assumed to be related, we have
(11) (3x) PA (x)
as an explanation of what it means for A to be permitted. Again, if the
same' relation is supposed to hold between our 'positive' obligation and
the notion of permission, the latter should be explained as
(12) P(x) A(x).
Neither (11) nor (12) seems to capture the meaning we are likely to
associate with the notion of permission. Saying that we are permitted to
do A ordinarily means something more than that we can in some situation
do A with impunity. Normally, some range of choice is implied in the
notion of permission. For instance, being permitted to drive a car means
more than that one may at least once drive a car without being liable to
punishment. Hence, (II) will not do. On the other hand, the following
example shows that (12) also leads into absurdity: In most countries, it is
legally permissible to consume alcoholic beverages; but it is absurd to
say that this permission is logically the same as a permission to drink on
every occasion. This, however, would be the case if (12) were an adequate
analysis of our notion of permission (as applied to kinds of acts); for
what is permitted by (12) is that every act under consideration is of the
SOME MAIN PROBLEMS OF DEONTIC LOGIC 6S
kind A. Hence, (12) will not do as a formulation of permission, either.

The most natural way of explicating the notion of permission seems
to be to say that acts of a certain kind are permitted if, in every particular
situation, one is allowed to perform an act of this kind. This means that
the most natural structure of our notion of permission is that of
(13) (x)PA(x).
It will tum out that (13) is neither equivalent to (11) nor to (12). Hence
the relation of permissions to obligations - like the relation of 'positive'
obligations to prohibitions - is more complicated than previously has
been assumed. I conclude, therefore, that these relations cannot be ex-
plained satisfactorily without quantifying words like 'every' and 'some'.
6. There is another group of reasons for employing act-individuals and

quantifiers in deontic logic. These reasons are derived from the problem
of interpreting the formulae that are built by means of' '" " '&', and 'v'
out of components some of which contain deontic operators and some
of which do not contain them. Formulae of this kind are useful for many
pertinent purposes. A representative example is offered to us by A. N.
Prior (see Formal Logic, Oarendon Press, Oxford, 1957, p. 221ff.). His
notation differs from ours in that he uses the letters a, b, ... for properties
of acts and not for individual acts. In his notation, they serve as a basis
of deontic operations which give us formulae like Po, 0", b, P(a&b),
and O(b va). But they can also by themselves serve as building blocks
for such formulae as (a&b), '" a, b&P(a v b), and Oa v ,.., c. Virtually
the only direction Prior gives for interpreting formulae of the latter kind
is that 'a' is to be read: 'An act of the sort A is done'.
One may ask here: What is meant by a formula like (a&b)? Pre-
supposing for the moment that the system under consideration is in no
way partial to any particular occasion, (a&b) may mean either (i) that
an act of the kind A is done and that (perhaps at some other time and on
some other occasion) an act of the kind B is done; or (ii) that on some
occasion or another an act with both the property A and the property B
is performed. Similarly, ,.., a may mean either (i) that no act of the kind
A is done or (ii) that an act of the kind ,.., A is done (i.e. that A is omitted)
at some time or another.
Both interpretations give rise to trouble. Irrespective of the choice be-
66 JAAKKO HINTIKKA
tween them, the other cannot be expressed within the system. Under the
interpretation (ii), many logical laws of the propositional calculus fail.
For instance, from a and '" a v b we can no longer infer b, from a and b
we cannot infer (a&b); etc.
Under the interpretation (i), these laws of the propositional calculus
are applicable. However, this interpretation has other undesirable conse-
quences. Under it, the meaning of the 'propositional' connectives '",',
'&', and 'v' is not the same within the scope of deontic operators and
outside of them. From this it follows that a formula of the form Pp may
mean something entirely different from saying that the state of affairs
expressed by p is permitted. For instance, under (i) '" a means that no
act of the kind A is done. Under the usual interpretation of deontic
operators, on the other hand, P", a does not mean that it is permissible
not to do any act of the kind A. It merely means that one is permitted
to do an act of the kind A, i.e. to omit A. Similarly, a &b means that an
act of the kind A is done and (independently of this) an act of the kind B
is done, while P(a&b) ordinarily means that one is permitted to do an
act with both the properties A and B.
The only way of obtaining a uniform interpretation of all our formulae
under the interpretation (i) is to change the meaning of formulae like
P(a&b) to say that one is permitted to do A and (independently of this)
to do B. Under this reading, formulae like P(a&b) == (Pa&Pb) should
be provable. They are not provable, however, in any current system. The
reinterpretation hence runs contrary to the intentions of the deontic logi-
cians. And for good reasons, indeed: the new interpretation would greatly
impoverish the applicability of deontic logic. Among other things, the
old interpretation is not expressible within a system in which the new
one is presupposed.
Thus, in order to obtain a satisfactory interpretation of Prior's for-
malism, we must choose the remaining alternative. We have to take
formulae like a&b to be concerned with some (arbitrarily chosen) partic-
ular situation. In other words, a&b should be read: 'An act of the sort A
is done in the particular situation we are considering, and an act of the
sort B is done in the same situation'. Under this interpretation alone is
one entitled to infer from 'An act of the sort A is done' and 'An act of
the sort B is done' that an act of both the sort A and the sort B is done.
This was seen to be indispensable if we want to obtain a uniform inter-
pretation of formulae like a v b v P(a&b) without sacrificing the rules

of propositional logic or the current interpretation of formulae like
P(a&b).
But when the limitation to one (arbitrarily chosen) situation is made
explicit, it also becomes obvious that the formulation cries out for gener-
alization. With this limitation, we cannot talk of what is done in general;
neither the interpretation (i) nor the interpretation (ii) can be expressed
within the system. We cannot even speak at one and the same time of
acts performed on two different occasions.
All these difficulties disappear if variables are introduced for individual
acts and quantification is permitted over these acts. For instance, the
two interpretations (i) and (ii) of Prior's (a&b) can now be explicated by
(3x)A(x)&(3x)B(x) and (3x)(A(x)&B(x». Similarly, the two inter-
pretations of - a can be formulated by -(3x) A (x) and (3 x) - A (x).
The limitation to a particular situation can be performed by explicating
a by A (a). An interpretation of Prior's formalism can then be obtained
within our system by assuming that one and the same free variable (and
no other one) occurs in all the formulae we consider simultaneously. This,
however, is only a part of our formalism. In the latter, we have formulae
like A (a) v B(b) which are concerned with acts done on different occa-
sions.
If we introduce variables for individual acts and employ quantifiers,
there is also no objection to assuming that the relations between '0', 'F',
and 'P' indicated by (1)-(6) hold in our system.
II. THE BASIC SEMANTICS OF NORMS
7. The above remarks have - hopefully - demonstrated the desirability

of employing act-individuals and quantifiers in deontic logic. There re-
mains the task of finding rules by means of which we can actually handle
quantifiers and other logical constants in the context of deontic operators
and, for that matter, handle deontic operators in general.
One may try to accomplish this by taking one of the current systems
of axioms and rules of inference for quantification theory and by looking
for suitable additional rules to govern the logiCal behavior of deontic
operators. I have found it advisable, however, to take a different line
of approach. In the procedure just mentioned, one easily loses touch with
68 JAAKKO HINTIKKA
interpretation and slips into assumptions which limit the generality of

one's theory. Likewise, not all the actually valid deontic principles can
easily be found in this way.
For these reasons, I shall proceed in a different manner. I shall first
try to define what it means for a (finite) set of formulae (with deontic
operators) to be satisfiable (consistent, logically possible). By means of
this notion, all the other most important notions can be defined: a formula
p is said to be satisfiable if and only if the set {p} whose only member is
p is satisfiable; p is said to be contradictory (logically false) if it is not
satisfiable; and p is said to be valid (logically true) if and only if '" pis
contradictory; q is a logical consequence of p (logically implied by p) if
and only if (p ::> q) is logically true, i.e. p & '" q not satisfiable. My ap-
proach can be called semantical or semi-semantical in contradistinction
to the older deductive (axiomatic) or 'syntactical' approach.
My treatment derives from an earlier treatment of quantification theory
along the same lines as well as from a similar semi-semantical theory of
modal logic. (See Jaakko Hintikka, Models for Modalities, D. Reidel,
Dordrecht, 1969.) Most of the formal considerations will turn out to be
special cases of this general theory of modal logic. For this reason I shall
omit some of the proofs, and concentrate on the aspects which are im-
portimt for applications.
8. What can we mean by saying that a set A. of formulae is satisfiable?

For the formulae of quantification theory pure, an answer that suits our
purposes can be given by starting from the (informal) idea that satisfiability
means capability of being true under some state of affairs. This answer
amounts to saying that A. is satisfiable if and only if it can be imbedded
in a set Jl which satisfies the following conditions:
(C. "') If p E Jl, then not", p E Jl.
(C.&) If (p & q) E Jl, then p E Jl and q E p.
(C.v) If (p v q) E Jl, then p E Jl or q E Jl.
(C.E) If (3x)p E Jl, thenp(a/x) E Jl for some free variable 'a'.
(C.U) If (x) p E Jl, then p (b/x) E Jl for every free variable 'b' occur-
ring in the formulae of Jl.
These conditions may be thought of as expressing properties of 'a
partial description of some possible course of events or some particular
state of affairs'. Although partial, this description nevertheless has to be

large enough to guarantee the possibility of the state of affairs or course
of events it describes.
We shall ourselves make use of this heuristic idea in the sequel. For
instance, (C. v) may be thought of as saying that if a disjunction is true
under some particular state of affairs, then at least one of its members
must be true under the same state of affairs. In quantification theory, the
conditions (C) are said to define a model set. In other words, 'model set'
is intended to be a formal counterpart to the idea of a partial description
of a 'possible world'.
In formulating the rules (C), we have assumed (for simplicity) that all
connectives other than ',..,', '&', or 'v' have been eliminated. We have
also assumed that all negation-signs have been distributed so as to precede
immediately 'atomic' parts of formulae (i.e. smallest consecutive parts
of formulae which are themselves formulae or like formulae except for
containing bound variables instead of free ones). This can always be
accomplished, for we can always drive each particular negation sign one
step deeper into the formula in question (or, perhaps, eliminate it alto-
gether) by means of one of the following laws: (a) de Morgan's laws;
(b) the law of double negation; (c) the equivalence of ,.., (3x)p with
(x) ,.., p and of ,.., (x) p with (3 x) - p. By applying these laws repeatedly,
we can ultimately arrive at a situation where our assumption concerning
the negation-signs is fulfilled. The same assumption can be made in
deontic logic, for we can here drive each negation-sign preceding '0' or
'P' one step deeper by applying to it one of the laws illustrated by (3)
and (4). ('F' may be assumed to be eliminated.)
9. The main problem is of course how formulae of the form Op and of

the form Pp affect the satisfiability of a set of formulae. In order to solve
this problem, we have to find rules supplementary to (C.-)-(C.U) which
will serve to take care of formulae containing deontic operators.
Let us first consider a sentence of the form Pp. What precisely is the
import of this sentence? What can bring out the fact thatp is permitted?
It is obvious that there need not be anything in the actual world which
suffices to show that p is permitted (permissible). (Permissions are not
always made use of; they are not betrayed by what people actually do or
fail to do.) When we are speaking of a permission obtaining in a world,
70 JAAKKO HINTIKKA
we are not speaking of this world only. Rather, we are speaking in some-
thing like a counterfactual or sUbjective mode. We are speaking of what
could be or could happen: we are saying that something could be done or
could be the case without violating any obligations (norms). In other
words, in saying that p is permissible in a world we are therefore saying
that this world can consistently be thought of as being replaced by another
in whichp is the case but in which all obligations are nevertheless fulfilled.
This implies that we cannot formulate conditions on the presence of
Pp in a description Jl of a possible world in the sole terms of Jl alone. In
addition to Jl, we have to consider another description Jl* of a possible
world, related to Jl in a certain way. This way will be expressed by saying
that Jl* is a deontic alternative to Jl. Intuitively, we may think of Jl* as a
description of that state of affairs in which p was assumed to be the case
for the purpose of showing that it can be the case while all obligations
are fulfilled. (Sometimes we shall also speak of the possible world de-
scribed by Jl* as an alternative to the world described by Jl.)
These observations motivate the adoption of the following condition:
(C.P*) If PpE Jl, then for at least one deontic alternative Jl* to Jl we
must have p E Jl*.
It -is, of course, assumed that deontic alternatives to a model set are
also model sets, i.e. that they satisfy the same conditions (C. '" )-(c. U)
as Jl.
This condition suffices to make sure that p can be the case, i.e. that there
is no inconsistency in assuming that p. But we wanted to accomplish
much more: we wanted to require that p can be the case while all obli-
gations are fulfilled. If this is to be accomplished by means of the deontic
alternative Jl*, we have to impose further conditions on it. We have to
require that whatever ought to be the case in the world described by Jl is
the case in the world specified by Jl*. In other words, we have to adopt the
following condition:
(C.O+) If Op E Jl and if Jl* is a deontic alternative to Jl, then p E Jl*.
This may already look like an exhaustive formulation of the intuitions

which underlie our concept of permission (permissibility) and obligation.
However, the handy formula 'p is permissible only if it can be the case
while all obligations are fulfilled' is subject to a subtle ambiguity. This
ambiguity can be brought out by asking: All obligations obtaining in

which world? We must not forget that we are here dealing with two
possible worlds, viz. those described by fl and fl*, respectively. Once this
is made explicit, we can at once see that further conditions are needed on
deontic alternatives. All the obligations that are required by (C.O+) to
be fulfilled are those that obtained in the 'old' world described by fl.
These are not all the relevant ones, however. Often we are only permitted
to do something at the expense of new obligations. If fl* is going to
demonstrate that p can be the case while all the relevant obligations
are fulfilled, these 'new' obligations have to be heeded, too. More
generally, we have to require that all the obligations obtaining in the
world described by fl* are fulfilled in it:
(C.O)r•• t If Op E fl* and if fl* is a deontic alternative to some model
set fl, then P E fl* .
Another, more general way of motivating the condition (C.O)rest is the
following. A deontic alternative to a given world, say the one described
by fl, is intended to be a kind of deontically perfect world, when viewed
upon from the point of view of the norms specified by fl. This require-
ment of deontic perfection means that all the relevant duties are fulfilled
in the alternative world. They naturally include all duties obtaining (in
virtue of the normative system specified by fl) in the alternative itself, no
matter whether they obtain in the world described by fl or not. These
duties are, of course, based on the normative principles in fl, but they
need not themselves be duties obtaining in the world described by fl. A
case in point might be a duty incurred by making use of a permission
that obtains in fl, a permission not itself utilized in the world fl speaks of.
Furthermore, it appears very odd to try to say that not all obligations
(norms) obtaining in the actual world obtain in its deontic alternatives.
Hence we have to assume that they do hold in these alternative possible
worlds, i.e. to adopt the following conditions:
(C.OO+) If Op Efland if fl* is a deonticalternativeto fl, then Op E fl*.
It is readily seen that (C.O)rest and (C.OO+) imply (C.O+). Hence this
last condition can be thought of as a derived requirement.
In view ofthe lesser obviousness of(C.OO+) as compared with (C.O+),
I have nevertheless listed the latter as a separate condition.
72 JAAKKO HINTIKKA
Finally, it seems reasonable to require that whatever is obligatory is

permissible. This is essentially the import of the following condition:
(C.o*) If Op E JJ., then for at least one deontic alternative JJ.* to JJ. we
must have p E p.*.
10. The main complication we have to take care of here is that the set JJ.*
mentioned in our conditions may in turn contain formulae of the forms
Pp and Op. In order to be sure that JJ.* is satisfiable we must hence apply
our conditions to p.* and not only to JJ.. The rule (C.P*) then compels us
to consider a third set p.** of formulae which is a deontic alternative to
JJ.* but not necessarily to JJ.. Again we may find it necessary to apply the
same condition to JJ.** so as to necessitate the consideration of still more
sets of formulae, etc.
For these reasons, we must define the notion of satisfiability in terms
of a set Q of sets instead of just two sets. Combining all the above obser-
vations, we may say that a set A of formulae is satisfiable if and only if it
can be imbedded in a member of a model system. A set Q of sets is a model
system if and only if it satisfies the following conditions: (I) every member
of Q satisfies the conditions (C.-)-(C.U); (II) a dyadic (two-place)
relation (calle.d the relation of deontic alternativeness) may be defined
among the members of Q in such a way that (i) each member JJ. of Q
satisfies (C.P*) and (C.o*) with respect to some member JJ.* of Q: (ii) any
two members JJ., JJ.* of Q satisfy (C.O+) and (C.OO+); (iii) each member
JJ.* of Q satisfies (C.O)rest if JJ. is in Q.
Further conditions (over and above those used to define a model
system) do not seem to be forthcoming. For instance, permissions ('per-
missibilities') cannot in any case be 'moved over' from JJ. to one of its
alternatives, say A, in the way (C.OO+) says that norms are 'movable',
for obviously one can quash a permission by making use of another.
It is important to realize that, and why, we often have to consider more
than one deontic alternative to a given world, say M. This is due to the
fact that not all permissions that obtain in M can be made use of in the
same world. The simplest case in point is a deontically neutral proposition
p, i.e. one for which PpEP., P - pEJJ.. Since p and - p cannot both be the
case in the same possible world, (C.P*) forces us to consider more than
one deontic alternative to JJ..
In formulating the semi-semantical conditions (C. '" )-{c.o*) above we

did not have to consider explicitly iterated deontic operators (deontic
operators occurring within the scope of other deontic operators). Conse-
quently, whatever reasons there may be for adopting these conditions,
they are independent of the problem of interpreting iterated deontic
operators. For this reason, it is all the more remarkable that the condi-
tions we have already accepted in effect give us a way of interpreting
sentences containing iterated deontic modalities, simply applying to them
the same principles that apply to noniterated ones. The controversies and
problems concerning iterated deontic operators can thus be by-passed
completely. In my judgement they merely illustrate the futility of attacking
interpretational problems by trying to formalize ordinary language
directly, without first developing suitable semantical tools which would
show precisely how the formalization is to be undertaken.
11. The semantical or semi-semantical conditions which we have just

explained are so central that they merit a number of further comments.
They are based on an idea which in some form or other occasionally
crops up in traditional moral philosophy, albeit often in a rather con-
fused form.
Perhaps the most important version of this idea (or a group of ideas)
is the notion of a 'Kingdom of Ends' (Reich der Zwecke) which we find
in Kant (See Grundlegung zur Metaphysik der Sitten. Our references are
to the second edition, whose pagination is usually indicated in critical
editions and translations.) It is occasionally characterized by him as a
'mere ideal' ('freilich nur ein Ideal') which is not realized but which we
nevertheless must be able to think of consistently. This state of affairs
would be realized, according to Kant, if all the maxims based on the
categorical imperative were followed without exception (op. cit., p. 438,
Kant's italics).
Since all moral maxims are for Kant based on the categorjcal impera-
tive, we can thus simplify and generalize a little and say that for us the
'Kingdom of Ends' is the world such as it would be if all and sundry
rational beings always honored all their obligations (duties). In this
respect, a Kantian 'Kingdom of Ends' is like a deontic alternative to the
actual world. These deontic alternatives are also 'deontically perfect
worlds' of sorts: all obligations, both those that obtain in the actual world
74 JAAKKO HINTIKKA
and those that would obtain in such an alternative POSt ible world, are
assumed to be fulfilled in each of them.
Notice also that for Kant the categorical imperative is obviously the
principle of all maxims, both of those obtaining in the actual world and
of those that are followed in a 'Kingdom of Ends'. Thus the requirement
that all these maxims are followed presumably implies that in a Kantian
'deontically perfect world' both the 'old' and the 'new' obligations are
fulfilled, just as our conditions (C.O·) and (C.O)rest require.
Generally speaking, the deontic alternatives to a given world are re-
lated to it rather in the same ways as a Kantian 'Kingdom of Ends' is
related to the actual world. From the point of view of this given world,
they are realizations of the (normative) ideals obtaining in it somewhat
in the same way as a Reich der Zwecke. We can say of them the same as
was said by Kant of a notion closely related to that of a Kingdom of
Ends, namely of the notion of an intelligible world (Verstandeswelt):
"The concept of an intelligible world is therefore only a point of view
(Standpunkt) which the reason finds necessary to adopt in order to think
of itself as an active being" (urn sich selbst als praktisch zu denken ... op.
cit., p. 458, Kant's italics). We could perhaps bring Kant's formulation
a little closer to ours if we changed the last few words to read: in order
to be able to think of itself as acting according to its normative principles.
Once Kant says that in morality a possible world of ends is considered as
if it were the actual world of nature (op. cit., p. 436, footnote).
In a sense the notion of a deontic alternative therefore is thus a some-
what watered-down and relativized variant of the Kantian notion of a
'Kingdom of Ends'. It is a much weaker notion because our concept does
not contain any reference to a particular moral principle, be it the cate-
gorical imperative, universalizability, or what not, in the way Kant's
notion does. It is a relativized notion, for it refers to the possible world
which a deontic alternative is alternative to (from the point of view of
which it is so to speak considered). Moreover, a deontic alternative to a
given possible world is not a unique entity, contrary to the way in which
Kant seems to have looked upon his 'Kingdom of Ends'. On the contrary,
normally there are several deontic alternatives to a given possible world
in a model system. This difference between us and Kant reflects partly
the more important role which is played by the notion of permission in
our thinking as compared with Kant. For it is precisely the multiplicity
of permissions that typically leads us to consider more than one deontic

alternative, as was pointed out above.
12. What my semi-semantical approach gives us in the first place is a

method of showing that certain sets of formulae are satisfiable, viz. by
means of exhibiting a suitable set 0 of sets. This is not enough, however,
for all the relevant purposes. We also need a method by means of which
we can show that certain formulae are valid, i.e. a method of proof. This
task reduces to the task of showing that certain sets of formulae are not
satisfiable, for (as we have seen) p is valid if and only if { '" p} (i.e. the
set whose only member is '" p) is not satisfiable.
A useful method of this kind may be obtained in the following way:
Let us suppose we are given a finite set A. of formulae. We shall try to
construct a model system 0 for A. by starting from A. and by adjoining
new formulae one by one so as to obtain bigger and better approximations
0' to O. Every adjunction is to remove one violation of the closure con-
ditions (C.&)-(C.o*) which help to define a model system. An approxi-
mation toward a model system obtained in this way is what we shall call
a provisional model system. Formally, it is simply a set of sets of formulae
on which a two-place relation is defined. Sets bearing this relation to a
given member are called its provisional (deontic) alternatives. Heuristi-
cally speaking, provisional alternatives are normally intended to be
approximations to 'full-fledged' alternatives.
The important thing about provisional model systems is the way in
which we can improve on one so as to obtain a bigger and better approxi-
mation toward a model system. This process is governed by certain rules
which may be called rules of analysis, in short, A-rules. Let 0' be a
provisional model system obtained by means of a finite number of
adjunctions. The following rules then serve to govern the introduction
of new formulae:
(A. &)1 If JI. EO' and (p & q) E JI. but not p E JI., then we may adjoin
p to JI..
(A.&h If JI. E 0' and (p & q) E JI. but not q E JI., then we may adjoin
q to JI..
(A. v) If JI. E 0' and (p v q) E JI. but neither p E JI. nor q E JI., then we
may adjoin p or q to JI..
76 JAAKKO HINTIKKA
(A. E) If Il e Q' and (3 x) p E Il but not p (b/x) E Il f<'r any free vari-
able 'b', then we may adjoinp(a/x) to Il, provided that 'a' is
a free variable not occurring in the formulae of any member
of Q'.
(A.U) If Il E Q' and (x)p E Il but not p(b/x) e Il for some free vari-
able 'b' occurring in the formulae of Il, then we may adjoin
p (b/x) to Il.
(A.P*) If Il E Q' and Pp E Il but not p E Il* for any member Il* of Q'
which is a provisional alternative to Il, then we may adjoin
{p} to Q'.
We shall say that the new member {p} of Q' is a provisional (deontic)
alternative to Il. This relation of provisional alternativeness is supposed
not to be disturbed by adjunctions by the rules (A) to Il or to {p}. We
also assume that all the relations of provisional alternativeness obtaining
between the members of Q' are originally due to an application of the
rule (A.P*).
If Il E Q', Il* E Q', Op E Il but not p e Il*, and if Il* is a
provisional alternative to Il, then we may adjoin p to Il*.
(A.OO+) 'If Il e Q', Il* E Q', Op E Il but not Op E Il* and if Il* is a
provisional alternative to Il, then we may adjoin Op to Il*.
(A.O)rest If Il* E Q', Op E Il*, but not p E IJ.*, then we may adjoin p
to Il* provided that Il* is a provisional alternative to some
member Il of Q'.
(A.o*) If Il e Q' and Op E Il but not P E Il* for any provisional
alternative Il* to Il in Q', then we may adjoin {p} to Q' so
as to become a provisional alternative to Il.
The same remark pertains to (A.o*) as to (A.P*).
How can these rules be employed to show that A. is not satisfiable? Let
us suppose that we have arrived at some provisional model system, and
suppose further that some member Il of Q' violates (C.,....,). Then it is
obvious from the definition of satisfiability that we have no hope of
reaching a model set by adjoining new formulae to the members of Q'.
The violation is there to stay.
In building a provisional model system we have a choice between two

alternatives every time we apply the rule (A. V); under this rule, we can
adjoin either p or q. It may happen that all the alternative ways of building
approximations to a model set also turn out to violate the consistency
condition (C ...... ). If this happens, then it is easy to show that the set A.
from which we started has no model set, i.e. that it is contradictory. This
is our basic soundness theorem. Although it is easy to establish, we shall
not try to prove it here.
This, then, is our method of showing that certain sets of formulae
are contradictory: We start from the set of sets the only member of which
is the given set A. which we are trying to prove contradictory. We apply
the rules (A) repeatedly to this set of sets. Every time we apply the rule
(A. v) we consider both the alternative outcomes of the application, i.e.
we use both of them as a basis for further applications of the rules (A).
We try so to arrange the applications of these rules that all the different
ways of building larger and larger sets eventually lead to a situation in
which some member of the sets of sets we have reached violated (C ...... ),
i.e. contains a formula together with its negation. If this happens in all
the alternative ways of applying the rules (A), then we know that the
original set A. is contradictory. It is also known (although we have not
proved it here) that if A. is contradictory, then a situation of this kind can
somehow be reached. This is the central completeness result of our deontic
logic.
In formulating this method, we have only considered the alternative
ways of applying the rules (A) to which the rule (A. v) gives rise. There
is the rule (A.E), however, which also seems to give different results
depending on the selection of the free variable 'a'. Our omission of these
different alternatives can be justified by showing that the choice of 'a'
does not make any difference for the outcome of our attempts to show
that A. is contradictory. Any method of picking up 'a' will do as long as
the provisos of (A.E) are fulfilled.
III. DEONTIC CONSEQUENCE VS. LOGICAL CONSEQUENCE
13. By means of the methods we have developed we can discuss several

conceptual problems connected with normative notions. As a preparation
for these applications, we must first make an important distinction. It
78 JAAKKO HINTIKKA
may be approached by asking: What does the validity of" a statement of

the form (p::::> q) mean? According to our definitions above it means that
the statement (p &'" q) is not satisfiable. The intuitive meaning of this
fact can be expressed by saying that p cannot be realized without ipso
facto realizing q, too. When we are thinking of or discussing normative
matters, we often - unwittingly - slip into discussing something else.
Without noticing it, we concentrate our attention, not on what can or
cannot be realized, but rather on what can or cannot be realized without
violating any obligations. In the case at hand, we are tacitly considering
whether p can be realized without realizing q while all norms are satisfied.
In other words, we are interested in whether (p &'" q) can be true, not
in any old possible world, but in a deontically perfect world. This means
that we are considering, not the satisfiability of (p & '" q), but the satis-
fiability of P(p&- q), in other words, not the validity of (p::::> q), but
the validity of 0 (p ::::> q). If (and only if) the former sentence is valid q
is usually said to be logically implied by p (to be a logical consequence of
p). If (and only if) the latter sentence 0 (p ::::> q) is valid, we shall say that
q is deontically implied by p (is a deontic consequence of p).
The point of view which served to connect our concepts with Kant's
also offers us illustrations of what the notion of deontic consequence
amounts to. In a logical consequence, we are asking what the realization
of p entails in any arbitrary possible world. In a deontic consequence,
we are asking what the realization of p entails in a 'deontically perfect
world' or, in Kantian terms, in a 'Kingdom of Ends'. This formulation
suggests a general, albeit somewhat vague reason why our intuitions
frequently pertain to relations of logical consequence in the realm of
logical relations between norms (and between norms and facts). It is
frequently much easier to be categorical about how things ought to be,
i.e. how they would be in a 'deontically perfect world', than to figure out
the complex duties one as a matter of fact has in the actual world. Hence
one is likely to have firmer intuitions, too, about the former than about
the latter.
This distinction between deontic and logical consequence has been
overlooked by most students of deontic logic, although it seems to be
implicit in certain concepts that have been used in traditional moral
philosophy. This neglect is all the more fatal as it often makes a crucial
difference whether the intuitions we seem to possess about deontic
concepts and about their interrelations are to be formulated as deontic

or as logical implications. The former is the case far more often than
logicians have realized. If, in such cases, the intuitions in question are
nevertheless forced on the Procrustean bed of logical implications,
fallacies are bound to arise.
14. The literature of deontic logic offers instructive and amusing exam-
ples of such fallacies. For instance, the following plausible-looking prin-
ciple has been put forward:
(14) If we are obliged to do A, then if our doing A implies that
we ought to do B, we are obliged to do B.
This certainly looks like a "quite plain truth" of logic, and it was taken
to be one by A. N. Prior in the first edition of his Formal Logic. He
formulated it essentially as follows:
(15) (Op & (p ~ Oq)) ~ Oq.
Whatever obviousness may seem to accrue to (14) belongs in fact to the
validity of the corresponding deontic consequence, i.e. to the validity of
(16) O«Op & (p ~ Oq)) ~ Oq).
This is shown by the fact that (15) is not valid on the assumptions we
have made, i.e. that its negation
(17) 0p&(-pAOq)&P-q
is satisfiable. This satisfiability is shown by the model system which
consists of the following two model sets:
(18) {Op,(- p v Oq),P - q, - p,(Op&(- p v Oq)&P - q)}
(19) top, p, - q}
of which the latter is assumed to be a deontic alternative to the former.
That this set of sets of sentences satisfies the defining conditions of-a
model system can be verified by inspection. (For simplicity, it is assumed
here that (C.&) is extended so as to apply also to conjunctions with more
than two members.)
What is even more interesting, our model system brings out the reason
why (15) is not valid. Model set (18), which may be viewed as representing
80 JAAKKO HINTIKKA
the actual world, contains both the sentence Op and the sentence '" p.
Using the terms employed in (14), this means that we can escape our
obligation to do B simply by failing to carry out the earlier obligation to
do A. There is nothing logically impossible or even logically awkward
about such a course of events: lamentable as it may be, many of our
obligations in fact remain unfulfilled. This 'escape' can only be prevented
by requiring that all our obligations be fulfilled. But to require this is in
effect to understand (14) as expressing a deontic consequence rather than
a logical one, i.e. as expressing the validity of (16) and not of (15).
That we really have a deontic consequence here can be demonstrated
by showing that the negation of (16), i.e.
(20) p(Op & ('" p v Oq) & P '" q),
cannot occur in any member of model system. This can be accomplished
by trying to build such a model system step by step by means of our rules
of analysis, and by showing that all ways of doing so lead to a violation
of (C. "'). The following argument serves to establish this and incidentally
also serves to illustrate the way in which the rules of analysis can be used
to establish validity:
We shall use one and the same letter D for the ever-growing approxi-
mations to the model system to be built (i.e. for successive provisional
model systems), and the lower case greek letters p, v, ... for their ever-
growing members. At each stage, D, p, v, ... contain only those members
they have been specified to have in the arguments so far. The first line of
our reductive argument shows what formula we are constructing a model
system for:
(21) p(Op & ('" p v Oq) & P '" q) E P. ED.
(22) Op & ('" p v Oq) & P '" q EVE Q.
This is obtained from (21) in virtue of (A.P·). Here v is a provisional

deontic alternative to p in D. Furthermore we have from (22) by (A.&)
(23) Op E v
(24) ('" p v Oq) E v
(25) P", q E v, hence
(26) p E v.
Here (26) is obtained from (23) by (A.O)rest. This rule is applicable because
v is a provisional deontic alternative to JI..
From (24) we obtain by (A. v)
(27) - pE V
or
(28) Oq E V.
But (26) and (27) violate (C. -). Hence we can restrict our attention to
(28). But (28) and (25) likewise violate (C. -), showing (in virtue of the
soundness of our method) the inconsistency of (20). If you are hesitant
about applying (C. -) to statements containing deontic operators, you
may continue the argument as follows:
(29) - q E e
from (25) by means of (A.P*). Here eis a provisional deontic alternative
to v in Q. But then we have
(30) q E e
from (28) by means of (A.O*). Here (29) and (30) violate once again
(C. -), this time as applied to simpler - possible atomic - statements.
This negative outcome of all the different ways of constructing a world
system for (21) shows the desired validity of (16). In our approach to
deontic logic, the situation is thus the one I claimed it to be: (14) has to
be interpreted as expressing a deontic rather than a logical consequence.
Nor is the possibility of showing this restricted in any way by the peculiar-
ities of our approach. In Prior's old system, the impossibility of adopting
(15) as a valid logical principle is shown by the unnatural consequences
of an attempted adoption. Prior dedu<:cs from this assumption a version
of the so-called first paradox of commitment. This is not the only awk-
ward consequence, however, nor the most awkward one. By substituting
(p&- p) for q in (15) and by noticing the disprovability of O(p&- p)
in most systems in deontic logic, including Prior's, we can readily deduce
from (15) the striking theorem
(31) Op => p.
This says that all obligations are in fact fulfilled, i.e. that we are dealing
82 1AAKKO HINTIKKA
with a 'deontically perfect world', precisely as I argued that we must in

the first place assume that we are doing.
In the second edition of Formal Logic (1962), Prior has given up (15),
prompted by syntactical considerations of the kind just mentioned. (He
attributes them to A. R. Anderson instead of my 1957 paper 'Quantifiers
in Deontic Logic' where they first appeared.) This does not seem to bring
out the full generality of the problem, however.
Prior is not the only logician who has been seduced by the usual
syntactical (axiomatic and deductive) methods into formulating perfectly
valid relations of deontic consequence as relations of logical consequence,
with all the absurd consequences resulting from such a course. It was a
rather inauspicious beginning for deontic logic that the very first axiom
of its very first attempted axiomatization embodied this mistake. This
axiomatization was offered by Ernst Mally in his monograph Grund-
gesetze des Sollens: Elemente der Logik des Willens, Graz 1926. His first
axiom was essentially the following:
(32) «(p::::> Oq) &(q ::::> r»)::::> (p::::> Or).
This may be criticized along the same lines as (15) was criticized above.
Again one can 'escape' the obligation that r simply by failing to carry
out the duty expressed by Oq. Again the fallacy involved here is illus-
trated by the strange consequences of the adoption of (32) as a logically
valid principle. Mally himself in effect deduced from his axioms the same
principle (31) as was seen to follow from the adoption of (15). This
consequence might seem counter-intuitive enough to overthrow any
axiomatization of deontic logic. Unfortunately, Mally was not deterred
by this strange consequence of his axioms. Apparently he found his
axioms so obvious as to be above suspicion. Instead, he resorted to the
hopeless expedient of trying to explain away the absurdity of (31) on
interpretational grounds. No wonder the subject was not carried further
for a while after Mally.
What makes (32) seductive is the fact that r is a deontic consequence of
(p::::> Oq), (q::::> r), and p. This can be verified without any trouble along
the same lines as the validity of (16) was shown above. Likewise, a
counter-example is easily constructed to shoo that (32) is not valid on
the basis of the assumptions we have made in the present paper.
Another example is provided by a principle which was put forward by
K. Grelling in his article 'Zur Logik der Sollsatze', Unity of Science

Forum, January 1939, pp. 44-47, and which prima facie is (in the words
of Prior) "not without certain intuitive plausibility":
(33) If the doing of A and B jointly necessitates the doing of C,
then if we do A and are obliged to do B, we are obliged to do C.
If this is interpreted as a logical consequence, it is a non sequitur, for the
conclusion can be avoided simply by assuming that the obligation to do
B is not fulfilled. This can be blocked by requiring that we are dealing
with 'deontically perfect world', i.e. with a relation of deontic conse-
quence. In fact, although the straightforward formalization
(34) (p & q) => r) => (p & Oq) => Or)
of (33) is invalid, it can easily be shown that r is a deontic consequence
of the antecedent of (34) together with (p&Oq).
Some of the awkward consequences of the adoption of (34) as a
logically valid principle have been pointed out by Prior. To them one
can add the deduction of (31) from (34) by substituting (p&-p) for r,
- p for p, and p for q.
Some of these fallacies may be avoided by strengthening material im-
plications into stricter ones. (According to Prior, this has been suggested
by G. E. Hughes as a way out of difficulties here.) This cannot be done
in all cases, however. For instance, there seems to be no hope of repairing
the 'Grelling paradox' (33) along these lines. Hence this suggestion cannot
constitute satisfactory diagnosis of what is involved in them.
These examples illustrate forcefully the need to distinguish deontic im-
plications from logical ones.
15. Another application of the concept of deontic consequence brings us

closer to traditional discussions of moral philosophy. In such discussions,
a considerable role has been played by what is known as the principle
that 'ought implies can' or in the original German terms as the 'sollen-
konnen' principle. (For a discussion of this subject, see e.g. G. H. von
Wright, Norm and Action, Routledge and Kegan Paul, London, 1963,
pp. 108-116, 122-125.) As the name indicates, the question here is
whether all oughts are intrinsically canny, that is, whether each obligation
presupposes a possibility of fulfilling it. The discussions of this problem
84 JAAKKO HINTIKKA
one can find in the literature can scarcely be said to have resulted in any
kind of consensus. If we express the concept of possibility by the modal
operator M (and the associated concept of necessity by the operator N),
it lies close at hand to think of the problem as being concerned with the
logical status of statements of form
(35) Op ::> Mp.
It is easily seen that on the assumptions so far made (together with
certain unproblematic assumptions concerning the notions of necessity
and possibility alone, unrelated to deontic notions) (35) is not valid. Of
course it can be made valid by adopting some further principles concerning
the interplay between the notions of possibility and necessity and deontic
notions. The merits and demerits of such principles would require a
longer discussion than can be undertaken here. It nevertheless seems safe
to say that no obvious and uncontroversial principle is forthcoming on
the level at which we are here moving to restore the validity of (35). In
a context of a logical discussion, it therefore seems advisable not to try
to salvage the 'ought implies can' principle by means of additional
assumptions.
It is perhaps worth emphasizing that a particularly forceful type of
argu1Dent for some versions of the principle is inapplicable here. It is
often said that 'ought implies can' because a man cannot be blamed for
not doing what he cannot do. And if he cannot be blamed for not doing
something, he cannot be under an obligation to do it. Hence his being
under such an obligation presupposes that he can fulfill it.
Whatever the merits of this line of argument are, it is inapplicable here.
What we are dealing with in the present paper are impersonal norms
rather than duties or obligations that pertain to some particular person.
(Whether, and if so how, the latter can be analyzed in terms of the former
is a question which will not be taken up here.) But if so, the argument just
sketched for the 'ought implies can' principle falls outside the scope of
the present paper, too, for it trades essentially in obligations of some
particular person. (Only by so doing can the crucial notion of blame be
brought in. For no one in particular can be blamed for not fulfilling an
'impersonal' norm of the kind we are here dealing with.)
Hence it may appear that very little can be said about the 'sollen-
konnen' principle here. One simple point can nevertheless be made.
SOME MAIN PROBLEMS OF DEONTIC LOGIC 8S
Whatever the status of (3S) is, there is no problem about the status of a
closely related sentence. Even if it is the case that Mp is not a logical
consequence of Op, it is without any doubt a deontic consequence of the
latter. In other words, sentences of the form
(36) O(Op ~ Mp)
are valid already in virtue of the assumptions we have made, plus one
unproblematic assumption concerning the logical behaviour of the con-
cepts of necessity and possibility.
This can be shown by means of our rules of analysis. An argument to
this effect can be carried out by trying to imbed the negation of (36) in
a member of a model system. As in our earlier argument, we shall use
a for the successive provisional model systems obtained in the course of
the arguments, and I" for the successive approximations of that member
of 0 in which we want the negation of (36) to occur. The argument may
proceed as follows:
(37) P(Op&N-p)el"ea.
This just says what we want to construct our model system for. (In
bringing the negation of (36) to the form here displayed, we have made
use of the equivalence of "" Mp with N ,.., p.)
(38) (Op&N-p)ev.
This is obtained from (37) by means of (A.P*). Here v is a provisional
deontic alternative to I" in D.
(39) Op e v (from (38) by (A.&)l)
(40) N - p e v (from (38) by (A.&)2)
(41) pev.
This is obtained from (39) by means of (A.O)rnv which is applicable
because v is a provisional deontic alternative to 1".
(42) - p e v.
This is obtained from (40) by the scarcely disputable principle that what-
ever is necessarily true (in a given possible world) is true (there). But (41)
and (42) contradict (C. "'), completing the desired argument, and thus
86 JAAKKO HINTIKKA
establishing the validity of (36). This means establishing that Mp is a

deontic consequence of Op. It is important to realize that the argument
by means of which this was established does not in any way tum on
assumptions concerning the interplay of deontic concepts with the notions
of possibility and necessity.
Our result is in itself very simple, and may even appear trivial - after
it has been established. (It ought to be the case that all duties are fulfilled.
Hence it ought to be possible to fulfill them). Some additional interest
is in any case lent to our observations by the possibility that the 'sollen-
konnen' principle was perhaps right from the beginning intended, how-
ever dimly and inarticulately, as an expression of a deontic consequence
rather than a logical consequence.
The principle was brought to prominence in moral philosophy by
Kant. Hence we have to ask: how did he conceive of it? Kant's expla-
nations are not distinguished by their lucidity, but an unmistakable and
recurrent tum of thought in Kant is in any case a connection between
'ought implies can' principle and the concept of freedom. (See e.g.
Critique ofPure Reason A 807, Critique ofPractical Reason, 1st ed., p. 54.)
Moral freedom, for Kant, lies in the very fact that a man can act in the
way he ought to act. On the other hand, Kant tells his readers that a man
exercises this freedom in so far as he is a member of that noumenal world
to which he occasionally assimilates his 'Kingdom of Ends' and which
on any showing behaves like the latter. Thus the fact that the moral law
is followed in that possible world which Kant calls the 'Kingdom of
Ends' or the 'noumenal world' is for him a ground for claiming that it
is possible for a man to follow the moral law. But if this is the case,
Kant's principle obviously amounts to a deontic rather than to a logical
consequence. What he is saying is not so much that an obligation logically
implies a possibility to fulfill it, but rather that the necessity of being
able to think (if only as an Idee) of all our obligations as being fulfilled
in some one world (at least in the noumenal world or in the 'Kingdom
of Ends') shows the possibility of human freedom and hence the possi-
bility of acting in accordance with our duties. On the basis of our earlier
remarks, these other possible worlds may be compared to our deontic
alternatives, and the fulfillment of all relevant duties in them will corre-
spond to what (C.O)reat (and in part also (C.O+» required. It is hence
no accident that the very rule (A.O)rest which corresponds to this con-
dition (C.O)rest played an absolutely essential role in the above argument

(37)-(42) by means of which we demonstrated the possibility of inter-
preting the 'sollen-konnen' principle as expressing a valid deontic conse-
quence.
From this point of view, the obscurity of many of Kant's formulations
will be but another illustration of the difficulty of telling deontic impli-
cations from logical ones - a difficulty from which modern philosophers
have not been found exempt, either. Be this as it may, it seems to me
unmistakable that the whole trend of Kant's thinking in moral matters
strongly suggests interpreting his 'sollen-konnen' principle as a deontic
rather than logical consequence.
IV. THE CONCEPT OF COMMITMENT. APPLICATIONS AND FURTHER

PROBLEMS
16. An interesting further problem is posed by the notion of commitment.

What is - or can be - meant by saying that a certain fact or act (let it be
described by p) commits one to acting a certain way, described (say) by q?
The two obvious candidates for this role that can be expressed in our
formalism are the following:
(43) O(p :::> q)

and
(44) p :::> Oq.
Much ink has been spilled in discussing th~ relative merits of these two
explications. It has been spilled in vain, for the conclusion seems to me
inescapable that our commonplace notion of commitment is intrinsically
ambiguous between the two renderings (43) and (44) (plus, possibly, still
others).
Our semantical insights enable us to appreciate the difference between
(43) and (44). The former reconstruction in effect assimilates, in the
special use in which (43) is logically true, the notion of commitment to our
earlier notion of deontic consequence. On this interpretation p commits
us to q if it is impossible to realize p in a 'deontically perfect world' with-
out realizing q, too. Since it has already been seen that our logical
intuitions in the area of normative concepts often in effect pertain to
88 JAAKKO HINTIKKA
relations of deontic consequence, it may be expected thai: also our ideas

of commitment must often be spelled out in terms of (43) rather than in
terms of its rival (44).1
Moreover, from this point of view we can see that the notorious para-
doxes of (derived) commitment are but particular cases of the paradoxes
of implication, and hence devoid of any special interest for a student of
deontic logic. The paradoxes consist in pointing out that, if (43) is a
satisfactory analysis of commitment, a forbidden act commits one to
everything and that everything commits one to an obligatory act. In
other words,
(45) 0- P ::J O(p ::J q)
and
(46) Op ::J O(q ::J p)
are said to be valid - as they of course are in our approach. However, if
the validity of (45) and (46) is looked upon from the point of view of our
'deontically perfect worlds', the appearance of a paradox is considerably
diminished. In (45), it is true to say thatp cannot be realized in a deontic-
ally perfect world without realizing q because p cannot be so realized
simpliciter. In (46), q cannot be realized in a deontically perfect world
without reali2;ing p, for p has to be realized in any such perfect world in
the first place. Thus the 'paradoxes' lose their sting against our inter-
pretation (43), provided that we realize what precisely it contains. At
worst we have a residual feeling of awkwardness which can be traced to
the same sources as the usual 'paradoxes' of entailment (implication).
We might also look at the matter slightly differently. There is little
doubt that as long as p and q are normatively neutral (neither obligatory
nor forbidden), (43) catches one sense in which we all frequently speak
of commitment. One obvious reason why the notion of commitment is
often employed is to prevent our actual world from departing from a
deontically perfect world. If p is the case and if it commits us to q in the
sense (43), then the actual world will not match the standards of deontic
ideality unless q will also be the case. To avoid this is one major purpose
which the announcement and enforcement of commitments of form (43)
is calculated to serve.
Of course, when p and q are not neutral, this purpose may become
otiose: if p is forbidden, a discrepancy between the actual world and
deontically perfect worlds has opened as soon as p has been realized,

irrespective of whether q is realized or not, and likewise for the case in
which q is obligatory. When the notion of commitment is used in such
unusual circumstances, it cannot usefully serve the purpose just indicated,
if construed as in (43). If the notion is nevertheless seriously employed
in such circumstances, some other purpose and hence some other inter-
pretation must be presumed. Thus on our analysis the paradoxes of
derived obligation (4S)-{46) do not show that the interpretation (43) of
commitment is misguided as an approximation of what is involved in
our idea of commitment in many ordinary contexts. At most they illus-
trate the fact that in those unusual circumstances with which (4S)-{46)
deal some notion of commitment different from (43) is tacitly presupposed.
17. This does not show, however, that (43) is always what people's
informal verbal statements about commitments presuppose even in
perfectly normal circumstances. In fact, there are good general reasons
for thinking that often (43) is not the intended interpretation. For one
thing, from (43) together with a purely factual statement no unconditional
statements of obligation follows. For instance, p and O(p::> q) do not
imply Oq. In this sense, commitments of the kind (43) do not admit
'detachment'. Yet on some occasions we certainly consider ourselves
justified to carry out such a detachment and to announce, on the basis
of a fact and a commitment, a definite non-conditional obligation.
A commitment for which this is possible must have something like the
force of (44) rather than (43). Whenever an actual obligation follows from
a commitment plus certain facts, some rec')nstruction along the lines of
(44) rather than of (43) is thus presupposed. Such cases seem in fact to
be quite common.
Objections have been made to (44) as an interpretation of the notion
of commitment. For instance, it has been alleged that on this interpre-
tation the realization of whatever is not in fact realized 'commits' one
to everything, for
(47) ,.., p ::> (p ::> Oq)
is valid.
The fact is that what creates the appearance of a paradox here is not so
much the idea on which (44) is based as rather the desire to have some
90 JAAKKO HINTIKKA
stronger implicational tie between p and Oq than a matC"rial implication

in (44). It is certainly true that in many of people's everyday uses of the
notion of commitment such a stronger tie is presupposed. However, it is
not clear to what extent this presupposition is due to 'pragmatic' or
'conversational' implications rather than to the basic logical force of the
expressions involved. In any case, the question concerning the nature of
this stronger tie is independent of our study of deontic notions which is
very well served - at least up to a point - by the simple material-impli-
cation explication (44).
18. The differences between (43) and (44) - as well as the reasons for using
both of them as alternative explications of the notion of commitment -
are illustrated by conflicts of duty. Such a conflict may e.g. result from a
promise. I give an honest promise (let this act be described by p) to bring
it about that q (say, have a cup of coffee with you). Unknownst to me,
however, my father has fallen ill, which creates an obligation to see him
that overrules my earlier promise. It seems to me that moral philosophers
have felt somewhat uncomfortable in discussing this kind of situation,
and perhaps one can also see why. For the fact that the obligation created
by my promise is overruled means that it is false to say simpliciter that
Oq, i.e. that I. am obliged to fulfill the promise. (By the same token, the
commitment involved in my promise cannot be construed as (44)). Yet
it is clear that not everything is morally all right if I have to break my
promise, however firmly this particular course of action may be pre-
scribed to me by the norms I abide by. I have somehow done something
wrong. This 'moral failure' is what easily makes one hesitant to say that
in such a case there is no absolute duty to keep the promise.
Our distinction between (43) and (44) enables us to see precisely what
goes wrong in such a case. It is obviously and clearly true, even in the
case of a promise overruled, that in a deontical/y perfect world such a
promise cannot be given without keeping it. In such a world, p cannot be
realized without bringing it about that q. Even if the act of promising
does not give rise to an actual duty to keep the promise (e.g. because of
other duties), it none the less remains true that in this sense giving a
promise commits one to keeping it.
The sense of commitment involved here is clearly (43). Thus it may be
said that we need sense (43) to account for the possibility that a perfectly
genuine commitment (e.g. a valid promise) may be overruled by other

obligations, while (44) is needed to do justice to the conceptual fact that
it sometimes does result in actual duties.
19. By this time the reader has - hopefully - begun to appreciate the
difference between (43) and (44). At the same time, an especially attentive
reader may also have had an experience of deja vu - of recognizing some-
thing he recalls from the literature of moral philosophy.
In fact, I have already slipped a few times into a bit of conventional
jargon by speaking in connection with (44) of actual or absolute obli-
gations. To make uninhibited use of this jargon, the contrast between (43)
and (44) is essentially that between prima facie duties (obligations) and
actual (absolute or overall) duties or obligations. 2
It is in fact obvious that the situations we considered for the purpose
of illustrating the difference between (43) and (44) are of the same kind
as those which the perpetrators of the traditional dichotomy have used
as paradigm cases of the contrast between prima facie duty and actual
duty. The main problem to which they have addressed themselves is
likewise admirably accounted for by our distinction. This problem is the
question as to how an obligation can be overruled and yet remain - in
some perfectly good sense - a genuine obligation. Our answer to this
question was already given. It is now seen to admit of a formulation in
terms of the traditional distinction.
In order to obtain an explicit reconstruction of the distinction prima
facie obligation vs. actual obligation, let us consider some set of normative
principles whose conjunction is n. (The sentences formulating these
principles may be of the form Oq, but they may also exemplify such more
complex forms as (43) or (44).) Let us also assume that we have as our
factual premises a set of descriptive statements whose conjunction is p.
Then we shall say that on the basis of the set of norms n, q is a prima
facie obligation if and only if
(48) 0 [(n & p) ~ q]

is valid, i.e. if and only if q is a deontic consequence of (n&p). Likewise,
there is (by definition) an actual obligation that q if and only if
(49) (n & p) ~ Oq
92 JAAKKO HINTIKKA
is valid, i.e. if and only if Oq is a logical consequence of ('I &p). Thus the
distinction between prima facie obligations and actual obligations is
closely related to the distinction between the notions of deontic conse-
quence and logical consequence. The ambiguity of our intuitions vis-a-vis
this distinction is probably the major reason why the notions of prima
facie obligation and actual obligation have been distinguished so late
and confused so often. (For a discussion of an example of such confusion,
see Sections 22 and 23 of the present essay.)
A logician is amused to find that an important philosophical distinction
once again turns out to be based on an operator-switch, i.e. on the
order of two different logical operations, in the case at hand, between
o and ::J.
Sir David Ross, who more than anyone else has been instrumental in
introducing the concept of prima facie duty (obligation) into contempo-
rary moral philosophy, uses the term absolute duty instead of actual duty
for its contrary. This is not incompatible with our reconstruction. The
fact that in (48) the deontic operator 0 governs a conditional (if-then)
sentence shows in what sense prima facie obligations are in our view non-
absolute or conditional.
If n does not contain any normative notions, there is a prima facie
obligation tha.t q if and only if q follows logically from the non-normative
premise (n&p). This fact throws some light on the notion of a technical
norm and on its relation to other kinds of norms.
20. A comparison with the usual explanations of the prima facie-actual

distinction readily shows the close relationship of this traditional dis-
tinction to our reconstruction, although I am perfectly willing to admit
that the traditional distinction has occasionally been put to uses which
our reconstruction does not catch. If anything, it seems to me that some
of the traditional moralists have been somewhat timid in following up
the implications of the distinction. Even Sir David Ross, in giving
examples of the failure ofprima facie duties to give rise to actual (absolute)
duties, does not emphasize strongly enough how often - and how
easily - such prima facie duties as e.g. arise out of a promise can, qua
actual duties, be overruled by other obligations. What makes moralists
hesitant to say, in the case of a failure of this kind, that no actual duty
obtains is undoubtedly the vague feeling that something goes morally
wrong in such cases. We have already seen, however, that this feeling is
sufficiently accounted for by pointing out the precise sense in which a
primafacie duty is violated: something takes place that would not happen
in a deontically perfect world. Indeed the actual breach of morality which
takes place in such cases is typically different from a failure to satisfy a
prima facie obligation. For instance, in the case of (say) promising the
only conclusion we can detach (the only actual duty we can infer) is the
actual duty not to give the kind of promise that will be overruled by
other obligations, assuming that the duty to fulfill a promise is a prima
facie one. In other words: although fromp and O(p::.> q) we cannot infer
Oq, we can from p, 0 (p ::.> q), and "" Oq infer 0 "" p.
Toward the end of this essay we shall meet one more instance of a
philosopher's failure to see how easily prima facie obligations can obtain
without any corresponding actual (absolute) obligation obtaining, and
without anything going wrong with our logic.
A reason for the importance of prima facie obligations follows from
our earlier remark that we are likely to have firmer views concerning
how things ought to be, i.e. what a deontica1ly perfect world looks like,
than concerning the multiple interrelations of actual duties. For what
prima facie obligations specify is precisely what happens in deontically
perfect worlds.
A bonus we obtain as a by-product of our reconstruction of the distinc-
tion between prima facie duties and actual duties is a handy terminology
for the distinction between the two kinds of commitment (43) and (44)
which was discussed above at length. The former may be called - and
from now on will be called - primafacie commitments and the latter actual
or absolute commitments.
21. Armed with these observations and distinctions, which help to

clarify the nature of commitment and the nature of prima facie duty, we
can approach what seems to me the prettiest fallacy (or group of fallacies)
one can find in recent philosophical discussion. This is the fallacy that
underlies John Searle's famous attempt to show 'How to Derive Ought
from Is'.3
Many of the details of Searle's subtle and suggestive paper are irrele-
vant to our concerns here. If we may simplify his main point a little,
Searle claims that from a purely factual premise (an 'is') describing an
94 JAAKKO HINTIKKA
act of promising together with the analytical (Searle uses the term
'tautological') premise that promises ought to be kept it follows that
there is an obligation to keep the promise in question (an 'ought'). In
short, an ought follows from an is plus a tautological and hence empty
additional premise.
Let p be a statement to the effect that a certain particular promise is
given and let q state that this particular promise is kept. We can relate
Searle's discussion to our own earlier discussion by expressing his
'tautological' second premise by saying that giving the promise in question
commits one to keeping it. Thus Searle's 'derivation of ought from is' is
(on our oversimplified reconstruction) of the form
P
(50) p commits one to q
Oq
Searle emphasizes that a promise is (analytically) an act of placing oneself

under an obligation to keep it. This undoubtedly brings out the analytical
(tautological) character of the second premise of (50) especially clearly.
However, it does not suffice to explain the precise logical form of the
argument (50).
To begin with, we shall not worry about the alleged analyticity of the
second premise of (50). The much more obvious trouble with (50) is its
ambiguity, due to the ambiguity of the notion of commitment. According
to what has been said earlier we have a choice between two readings of
(50):
P
(50*) O(p::::> q)
Oq
and
p
(50**) p::::> Oq.
Oq
This distinction between two senses of (50) corresponds neatly to the two
senses of Searle's locution 'placing oneself under an obligation'. In (50*)
the obligation in question is a prima facie obligation, while in (50**) it

is an absolute one.
22. How is Searle's argument to be evaluated in view of this ambiguity?

Earlier, it was hinted that perhaps the most common notion of commit-
ment is something like (43). Accordingly, we might expect that the most
plausible interpretation of Searle's argument is (50*). Unfortunately,
(50*) is not a valid inference. (This can be seen by constructing a model
system a member of which contains p, 0 ( '" p v q) and P", q - if the
point is not obvious enough.)
In contrast, (50**) is a valid inference, indeed an instance of modus
ponens. Does this show that Searle is right? No, it does not. This re-
construction is based on the assumption that the notion of obligation
involved in one's obligation to keep one's promises is an absolute (actual)
obligation. If we adopt this position, then it becomes dubious whether
the second premise (44) of (50), interpreted as (50**), is really analytical,
as Searle claims. At first blush, it certainly appears patently false to say
that an act of promising entails (analytically!) an absolute (actual,
overall) obligation to keep it. Saying this seems to overlook completely
the possibility that the prima facie obligation which is admittedly created
by the promise should be overruled by some perfectly valid competing
obligation. It was precisely to account for this possibility that absolute
(actual) obligations were distinguished from the prima faCie. ones in the
first place. But if so, the second premise of (50**) is not analytical (and
may in fact be contingently false).
Thus it seems that the second reconstrual (50**) of Searle's argument
fails as badly as the first one to serve the purpose it was calculated to
serve. Although it yields a formally correct piece of reasoning, the
resulting second premise is not analytical. Hence the 'ought' conclusion
does not follow from an 'is' alone, but only in connection With another
(non-tautological) 'ought'.
However, this is not the only possible way of viewing (50**). One might
try to insist, after all, that its second premise is analytical. Of course, this
stratagem will succeed only if the first premise p can somehow or other
be strengthened. This extra strength can be sought for in two different
directions. We may either require more of the notion of promising than
before, so much more indeed that (44) becomes analytical in the case at
96 JAAKKO HINTIKKA
hand. Alternatively, we may want to conjoin p (the statement that a

promise is given) with the statement (let us call it cp) that certain ceteris
paribus conditions are satisfied. As far as the formal structure of the argu-
ment is concerned, the two suggestions result in parallel treatments, the
second differing from the first only in that the role of p is now played by
the conjunction (p&cp).
Let us examine the first line of thought first. In this case, the truth of
p cannot any longer be ascertained simply by examining the person in
question, his actions, and his circumstances. However clearly he says
'I promise', however sincere he is, and however fully all the other factu-
ally ascertainable presuppositions of successful promise-giving are satis-
fied, we still cannot say in the intended sense that he promised (the in-
tended sense here being the one in which promising means shouldering
an absolute obligation to keep it) unless it can also be established that
there are no stronger competing obligations for him to fulfill. But
to show this is not to establish a matter of fact. It means evalua-
ting the overall normative situation so as to ascertain that a certain
prima facie duty is not overruled. In brief, the truth of p no longer
amounts to a matter of fact; p is no longer a factual premise but a
normative one.
On this interpretation, there is no formal fallacy in (50**). Moreover,
the second premise is analytical all right. However, now the first premise
p is not a descriptive statement, but contains a normative component.
Hence the argument again fails to provide us with a 'derivation of ought
from is': now the first premise is no longer an 'is'.
This line of interpretation is somewhat unrealistic in any case, for no
one is likely to maintain seriously as strong a notion of promising
('really' promising) as is required for it. No one is likely to maintain, that
is, that it is part and parcel even of some unusually strong sense of
promising that giving a promise means undertaking an absolute obliga-
tion to keep it. However, there is a much stronger temptation to attempt
the other way out and replace p by (p&cp), that is to say, to maintain
that although promising in itself does not entail an absolute duty to keep
the promise, it does so provided that certain ceteris paribus conditions
are satisfied. This does not make any difference, however, for then what
was just said of p will apply mutatis mutandis to the conjunction (p &cp).
For the reasons given, (p&cp) will not be a purely factual statement. If
p does not contain any normative elements, then the ceteris paribw
condition cp will be at least partly normative.
We might thus represent schematically the three interpretations of
(50**) which we have considered as follows:
p (factual)
(51) p => Oq (non-analytical)
Oq
p (normative)
(51 *) p => Oq (analytical)
Oq
(p & cp) (normative)
(51**) (p & cp) => Oq (analytical)
Oq
Although all these three represent logically valid inferences, they fail
to provide us with a 'derivation of ought from is' in the intended sense.
The specious plausibility of assuming, in the third line of interpretation
just mentioned, that the ceteris parihw condition cp can be taken to be
factual is witnessed by Searle's adherence (essentially) to this line of
defense. He formulates the second premise as follows: "All those who
[promise, i.e.] place themselves under an obligation are, other things
being equal, under an obligation." He explains the need of the qualifying
clause here by saying that "we need the ceteris paribus clause to eliminate
the possibility that something extraneous to the relation of 'obligation'
to 'ought' might interfere." The interfering factors that Searle here labels
'extraneous' include competing stronger obligations, which the ceteris
paribus clause must also eliminate in order to serve its purpose. But they
cannot be ruled out by means of purely factual assumptions.
Searle nevertheless strives to maintain that "there is nothing necessarily
evaluative about the ceteris paribus conditions." His argument hinges
on the observations that "an evaluation [of the competing obligations]
is not necessary in every case" and that "unless we have some reason to
the contrary, the ceteris paribw condition is satisfied, and the question
whether he ought to do it is settled by saying 'he promised'." This argu-
ment has no force, however. It is true that in some cases no intervention
98 lAAKKO HINTIKKA
takes place, i.e. that in some normative situations there are no conflicting
obligations. But to state that this is the case is to make a normative
statement. Saying that t is deontically neutral (,..., Ot &,..., 0 ,..., t) is as
much a normative statement as saying that it is obligatory or forbidden.
Likewise, to say that a prima facie obligation is not overruled by others
(and that the question of actual duty can be decided in the way Searle
says) is to make a normative statement, however negative. And to try to
tie the need of evaluation to the question whether counter-arguments
have in fact been presented, or whether we (actually?) have 'reasons to
the contrary', as Searle appears to do, is of course beside the point. The
question is what is implied by the norms one has accepted, not what
arguments have actually been put forward or what reasons one actually
has.
23. This by no means exhausts the interest of Searle's clever paper, nor
even the different types of argument he is considering. A closer exami-
nation of these alternative arguments would uncover flaws in them
similar to those we have already discussed. 4 However, our sole purpose
here is to illustrate such important distinctions as (43)-(44) by means of
Searle's main argument, which does not motivate a discussion of the
further details of his paper.
One can nevertheless use our distinctions also to emphasize the extent
to which Searle is perfectly right. If 0 (p ::> q) is a principle of one's
normative system, however analytical, then one can after all infer from
p that there is a prima facie obligation to bring it about that q. (To see
this, put O(p::> q) for n in (48) and try to assume that its negation is
satisfiable.) Hence Searle is right in a rather striking sense. He has in
effect pointed out that from an 'is' and from an analytical principle one
may legitimately derive a perfectly genuine obligation, viz. a prima facie
obligation. An 'ought' does follow from an 'is', albeit only a prima facie
'ought'. This observation becomes all the more important in the light of
our earlier observation that such a prima facie 'ought' is often what our
intuitions are all about anyway.
From this point of view, the basic flaw of Searle's paper does not
consist so much in putting forward a fallacious argument as in failing to
spell out the sense in which his (correct) conclusion is to be understood.
He is calling our attention to a perfectly legitimate relation of deontic
consequence but discussing it as if it were a logical consequence. His

argument thus illustrates once again a type of confusion which we have
already noted several times in the course of the present paper. 5
This does not completely spoil Searle's main purpose, however. Part of
his general emphasis I can in fact share whole-heartedly. This part is the
subtlety and multiplicity of the ways in which normative and factual
concepts are interrelated. What we have discovered in the present paper
tends to reinforce rather than to lessen this impression of interrelatedness.
I believe that this impression is fully justified, and that it will be strength-
ened by further studies. (For instance, such studies may be expected to
bring to light the large extent to which the meaning postulates of many
salient concepts contain both factual and normative elements.)
At the same time Searle's paper fails to establish another part of his
aim, provided that the results of our critical examination of Searle are
justified. If we are right, his argument does not show that we cannot
carry out a sharp distinction between 'ought' and 'is' in the sense that in
an appropriate explanatory model of our normative discourse the non-
descriptive element is compressed into the deontic operators 0 and P
and that the logical laws governing these operators will obey important
'conservation principles' reminiscent of Hume's dictum on 'ought' and
'is'. In these respects, our discussion is more likely to comfort Hume than
Searle. If these questions are to be emphasized, we shall have to say (I
have suggested) that an 'ought' does not follow from an 'is'.
24. This discussion of Searle was calculated to illustrate the general

conceptual framework developed in the second and third chapters
(Sections 7-13) of this work, and to bring out some of the implications of
my remarks on the notions of commitments and duty. What has been
said does not amount to an exhaustive analysis of the logical structure
of the notion of commitments, however.
One reason for saying this is that we have not yet considered that
quantificational structure of the notion of commitment whose importance
was emphasized above in Chapter I.
What is this structure - or, if there are several types of commitment,
what are these structures? In answering this question, we have to distin-
guish, of course, between absolute and prima facie commitments. What,
first of all, is the typical form of absolute commitments? What would be
100 JAAKKO HINTIKKA
meant by saying, for instance, that by borrowing money one incurs an

absolute obligation to pay it back?
As a rule, the best answer seems to be to say that in such cases each
case of borrowing implies an obligation that there be at least one act on
the part of the borrower which is an act of paying it back. If 'B(x)' means
that x is an act of borrowing money and if 'R{x, y)' says that y is an act
of repaying the debt incurred in the act x, then the logical form of the
absolute commitment in question will be
(52) (x) (B(x) :::> O(3y) R(x, y»).
In general, the typical form of absolute commitment seems to me to be
(53) (x)(p:::> O(3y)q).
Here 'x' occurs inp and normally both 'x' and 'y' occur in q. However,
absolute commitments of form (53) are not the only ones. In analogy to
the 'negative' obligations mentioned above in Section 5, there are also
absolute commitments of form
(54) (x)(p:::> O(y) q).
For instance, it is reasonable to say that becoming a naturalized citizen
of a country commits one to keeping its laws. By this we, of course, mean
an obligation to the effect that all of one's subsequent acts conform to the
laws of the country in question. Hence the form (54). However, to my
'moral sense' commitments of this form smack of prohibition. They
should be carefully distinguished from 'positive' commitments of form
(53).
Likewise we have two different kinds of prima facie commitments,
one of form
(55) O(x) [p :::> (3y) q],
the other of the form
(56) O(x) [p :::> (y) q].
In (55) and (56) the variables 'x' and 'y' both normally occur in q.
However, it seems to me that in all the four cases (53)-(54) and (55)-(56)
it is often possible to describe the act which fulfills the commitment
without explicitly mentioning the act through which the commitment was
made. Then q may be thought of as not containing any occurrences of x.
25. It was indicated earlier that our conditions and rules concerning
quantifiers are in need of modifications. In the context of the usual
modal operators, this need is pointed out in my paper, 'Modality and
Quantification', Theoria 27 (1961) 119-128 (reprinted with additions in
Modelslor Modalities, D. Reidel, Dordrecht, 1969). For details, it suffices
to refer the reader to these other discussions. The main point is clear
enough in any case. The conditions (C. U) and (C.E) obviously presuppose
that each individual whose name occurs in the members of p exists in the
possible world described by p. Now an individual existing in one world
need not exist in others. Hence rules which for the first time import a
new free individual variable into (an approximation toward) a model set
are suspect, and have to be modified so as to rule out illicit impor-
tation.
By inspection, it is seen that there are two such rules, viz. (A.O+) and
(A.OO+). They will have to be changed as follows:
(A.O*) Like (A.O+) except that it is also required that all the free
individual variables of p occur in the members of p*.
(A.OO*) Like (A.OO+) except that it is also required that all the free
individual variables of p occur in the members of p*.
Corresponding changes are of course needed in the conditions (C.O+)

and (C.OO+):
(C.O*) If Op E p, if p* is a deontic alternative to p, and if all the free

individual variables of p occur in the members of p*, then
PEP*·
(C.OO*) If Op E p, if p* is a deontic alternative to p, and if all the free
individual variables of p occur in the members of p*, then
OPEP*·
The individuation of acts is not a straightforward matter, and it would

take us too long to discuss it adequately here.· It can nevertheless be
shown, it seems to me, that the cautionary measures we just carried out
are needed also when our individuals are individual acts. As an example,
102 JAAKKO HINTIKKA
we may consider an implication of the following form:

(57) (3x) OA(x) => O(3x) A(x)
This is obviously unacceptable intuitively as a logically valid principle.
If there is, under the actual course of events, an act that ought to be an
instance of forgiving a trespass, it clearly does not follow that there ought
to be, under any deontically perfect course of events, an act of forgiving
- and hence, presumably, also another earlier act of trespassing.
Yet (57) can be 'proved' to be valid if our original conditions (C.O+)
and (C.OO+) are assumed. An argument to this effect is easily carried
out. It shows where the trouble lies, for one of its steps will move over a
free individual symbol for the act mentioned in the antecedent of (57)
from a description of the actual world to one of its deontic alternatives,
i.e. to the description of a deonticaIIy perfect world. However, there is
no reason why in such a perfect world there should exist any counterpart
to this act in the actual world, and every reason why there should not.
Hence in this case the use of (A.O+) and (AOO+) as distinguished from
(A.O*) and (A.OO*) is clearly illicit.
As I have pointed out elsewhele, in a purely deductive (syntactic)
treatment assumptions tantamount to (AO+) and (AOO+) (as distin-
guished from (A.O*) and (A.OO*» easily sneak in. The case with which
we have diagnosed the situation may therefore be considered an argument
for a semantical or semi-semantical treatment of the kind undertaken
here.
Examples of the same kind can be multiplied. For instance, from the
fact that there in fact has occurred an act (which therefore 'exists' in the
usual nontemporal sense of existence) which ought to be punished, it
does not follow that there ought to exist a punishable act. In other
symbols,
(58) (3x) O(3y) R(y, x) => o (3x)(3y) R(y, x)
is not valid after the changes indicated in the preceding section have been
made. Here R(y, x) may be read 'y is a punishment for x'.
In this way, some forms of the so-called paradox of the Good Samaritan
can be dealt with. Other conceptual problems in this direction - in
particular, certain forms of contrary-to-duty obligations - nevertheless
require an altogether different sort of treatment.
One more point deserves to be mentioned here. Accepting such stronger

conditions as (C.O+) and (C.OO+) means (very nearly) assuming that
actually existing acts can be 'moved' to deontic alternatives, i.e. that they
exist in all of them. If we make this assumption, it is hard to see why the
converse 'moving over' of individuals should not be allowed, too. This
would mean modifying (C.E) and (C.U) so as to allow as substitution-
instances of 'x' free individual variables which occur in the alternatives
to Jl but not in Jl itself. Then, and only then, would (7) imply (8), as one
can easily check.
However, none of the 'moving over' assumptions has the slightest
intuitive plausibility, and they should all be kept out of a realistic and
flexible deontic logic.
Among the main possibilities that remain to be investigated there are
the following:
(i) Systems combining deontic operators with operators for necessity
and possibility.
(ii) Systems in which the class of 'deontically perfect worlds' is
structured further, for instance, into more or less perfect ones, or into
worlds that are as perfect as certain boundary conditions allow. In a
recent paper, Bengt Hansson explores this last possibility in an interesting
way. (See Hansson's paper in this volume, p. 121-147.)
Other possibilities in this direction are stilI largely open.
Academy of Finland.
NOTES
1 The main difference between the distinction (43)-(45) and the earlier distinction,
deontic consequence vs. logical consequence, is of course that neither (43) nor (44)
has to be true for logical (conceptual reasons, whereas the latter distinction dealt with
two kinds of logical (conceptual) connections between statements.
2 The primary sources of this distinction in recent moral philosophy are the writings
of Sir David Ross, especially The Right and the Good, Clarendon Press, Oxford, 1930,
and The Foundations of Ethics, Clarendon Press, Oxford, 1939.
3 John R. Searle, 'How to Derive Ought from Is', Philosophical Review 73 (1964)
43-58, reprinted in Jerry H. Gill (ed.), Philosophy To-Day no. 1, The Macmillan Co.,
New York, 1968, pp. 218-235, and in Philippa Foot (ed.), Theories of Ethics, Oxford
Readings in Philosophy, Oxford University Press, Oxford, 1968, pp. 101-114.
4 A case in point is the following: Searle says that the kind of criticism I just presented
is in any case inconclusive, "for we can always rewrite the relevant steps ... so that
they include the ceteris paribus clause as part of the conclusion."
104 JAAKKO HINTIKKA
This sounds fine. However, everything depends on the precise way in which the
incorporation of the ceteris paribus condition in the conclusion IS supposed to be
accomplished. There are two possibilities which yield essentially the following putative
arguments:
p
(*) p O(cp ::> p)
::>
O(cp ::> p)
P
(**) p ::>(cp ::> Op)
(cp::> Oq)
Now the second premise of (*) is obviously false. Surely it does not follow from the
fact that a promise is given in the actual world that in all deontically perfect worlds cp
is followed by q. Hence (*) must be ruled out.
The only way to avoid this conclusion is to make cp ~ p a logical (analytical) truth.
This, however, deprives (*) of all relevance for Searle's purpose.
In (**), the conclusion is a conditional with tacitly normative antecedent and an
explicitly normative consequent. That such a statement follows logically from a factual
statement together with an analytically true premise has no implications whatsoever
for the is-ought distinction, any more than (say) the logical truth of Oq ::> Oq has.
5 In his new book, Speech Acts, Cambridge University Press, Cambridge, 1969, p. 181,
Searle now distinguishes between two kinds of obligations, exemplified by
All things considered, Jones ought to pay Smith five dollars
and
As regards his obligation to pay Smith five dollars, Jones ought to pay
Smith five dollars,
respectively. He says that only obligations of the latter type, not of the former, can be
derived from an 'is'. This distinction comes very close to our distinction between
absolute obligations and prima facie ones. Searle fails to spell out, however, precisely
what is involved in the latter. I claim that when this is done, the limitations of Searle's
argument become patent. A derivation of a prima facie ought from an 'is' does not
violate the fact-norm dichotomy, correctly understood.
GEORG HENRIK VON WRIGHT
A NEW SYSTEM OF DEONTIC LOGIC·
In a recent paper! Professor Roderick Chisholm has drawn attention

to an important normative idea which he calls "the contrary-to-duty
imperative". Chisholm criticizes some systems of so-called deontic logic,
among them the system proposed by me in 1951,2 on the ground that
they are unable to cope with this normative idea. He also makes some
positive suggestions concerning the logical structure of the imperatives
in question.
II
Before I proceed to a discussion of contrary-to-duty imperatives, I shall

present a formalized version of essentially the same system of deontic
logic as the one put forward in my early paper. I shall refer to this for-
malized theory as the Old System of Deontic Logic.
The system has the following vocabulary: a. An unlimited supply of
variables A, B, ... b. The truth-connectives -, &, v, --+, ~, for negation,
conjunction, disjunction, (material) implication, and (material) equiva-
lence respectively. c. An operator o. d. Brackets.
The variables I regard as schematic descriptions of a type of proposi-
tion-like entity which I propose to call 'generic states of affairs'. A generic
state of affairs may obtain on a certain occasion and may not obtain on
some other occasion. For example: that the window of my room is
closed is a generic state of affairs. On the occasion when I write this, the
window is closed. On many other occasions it is open. I shall not here
further try to elucidate the notion of a generic state of affairs nor that of
an occasion. 3
The well-formed formulae of the system I shall call O-expressions. By
an atomic O-expression I understand an expression formed of the letter
o followed by a variable or by a molecular complex of one or several
variables and truth-connectives. By O-expressions generally I understand
R. Hilpinen (ed.) , Deontic Logic: Introductory and Sy.tematic Reading., 105-120. All right. re.erved.
Copyright © 1970 by D. Reidel Publishing Company, Dordrecht-Holland.
106 GEORG HENRIK VON WRIGHT
atomic O-expressions and their molecular complexes. (T1:le rules for the
use of brackets in formulae I shall not mention here. Which they are
should be obvious from the presentation.)
The formula 'OA' may be read as follows: one ought to see to it that A.
And similarly for other formulae. For example: If 'A' describes the state
of affairs that the window is open, 'OA' says that one ought to see to it
that the window is open. - The reading of 'OA' as one ought to do A
I do no longer regard as fully correct.
The system has two axioms:
(AI) ,..., (OA & 0,..., A).
(A2) O(A & B) +-+ OA & OB.
The rules of inference are the following four:

RI For any variable in an axiom or theorem of the system may
be substituted (throughout) another variable or molecular
compound of variables.
R2 Modus ponens.
R3 A variable or molecular compound of variables in an axiom
or. theorem may become replaced by a tautologically equi-
valent compound of variables.
R4 The O-expression which is obtained from a tautology of
propositional logic by replacing its propositional variables
by O-expressions is a theorem.
The system is decidable. We need not here discuss a decision procedure
and other meta-logical features of the system.
III
Assume that an agent has a duty to see to it that A. Assume that he

neglects his duty, i.e. that he acts in such a manner that it comes to be
that ,..., A. With this possi bility in mind we may wish to lay down a norm
(imperative) to the effect that, if the agent neglects his duty to see to it
that A, then he has a duty to see to it that B. This norm is what Chisholm
calls a 'contrary-to-duty imperative'.
A NEW SYSTEM OF DEONTIC LOGIC 107
For example: Assume that it is our moral duty not to hurt anybody's
feelings. Assume that we have hurt somebody's feelings. Then our duty
could be to apologize. This, I understand, would be a contrary-to-duty
imperative in Chisholm's sense.
It is obvious that contrary-to-duty imperatives play an important rOle
in the moral life of man, and in law. 'Rules of reparation', i.e. norms
concerning how to make good some damage caused or some evil done,
are often of this kind. It is a definite merit of Professor Chisholm's to
have drawn attention to this type of imperative in the context of deontic
logic.
IV
Which is the logical structure of contrary-to-duty imperatives? How shall

we express them in a formal language? In particular: Can they be
'formalized' in the Old System of Deontic Logic?
Let 'OA' express a (primary) duty to see to it thatA. Could '0(,..., A -+ B)'
express the contrary-to-duty imperative to see to it that B in case we
neglect our primary duty? It is not difficult to see that the answer is
negative.
By virtue of R3 and R4, OA-O[(A v B)&(A v -B)] is a theorem.
By virtue of A2 and RI, OrcA v B)&(A v,..., B)]-O(A v B)&O(A v
,..., B) is a theorem. By virtue ofthe above and R2 and R4, OA -+ 0 (A v B)
is a theorem. By virtue of this and R3, finally, OA ~ 0 ("" A -+ B) is a
theorem.
Thus it may be proved that the primary duty entails the proposed
contrary-to-duty imperative. If '0 ( ,..., A -, B)' expresses a contrary-to-
duty imperative, it follows that, no matter what state of affairs 'B' de-
scribes, it would be our duty to see to it that B, had we neglected our
duty to see to it that A. This is absurd.
]n the Old System of Deontic Logic '0 ("" A -+ B)' is the only initially
plausible candidate for the job of expressing a contrary-to-duty imper-
ative (relative to a primary duty expressed by 'OA'). As we have seen,
and as Chisholm points out, the candidate cannot be accepted. The Old
System is incapable of expressing contrary-to-duty imperatives. Since
imperatives of this type are important in the normative life of men, this
incapacity is a serious shortcoming of the system. To have noted this is
another merit of Professor Chisholm's.
Chisholm also makes a constructive suggestion of his own concerning

the logical form of contrary-to-duty imperatives. According to his
suggestion, ' ...... A --+ OB' might be regarded as expressing a contrary-to-
duty imperative relative to a primary duty expressed by 'OA'.4
I do not think that this is a happy suggestion. I shall not, however,
here try to criticize it in detail. I am more anxious to present an alternative
suggestion in the matter which I consider more helpful. A few words of
criticism will therefore suffice.
, ...... A --+ 0 B' is not a well-formed formula of the Old System as we have
presented this system here. This observation, however, I shall not make
a ground of criticism of Chisholm's suggestion. We could build a calculus
which allows molecular compounds of O-expressions and variables and
compounds of variables. There are some difficulties in interpreting such
'mixed' expressions of 'factual' and 'normative' components. But these
difficulties can, I think, be overcome.
Let there be a primary duty to see to it that A. Assume that' ...... A --+ OB'
expressed the contrary-to-duty imperative to see to it that B in case of
neglect of the primary duty. Assume further that the duty-bound agent
fulfils his prim.ary duty, i.e. sees to it that A.
A --+ ( ...... A --+ OB) is a tautology. Thus, the assumption that the agent
fulfils his primary duty seems to entail 5 that, had he not fulfilled his duty,
he ought to have seen to it that B. This follows no matter what state of
affairs 'B' describes. Thus it also would follow that, had he not fulfilled
his duty, he ought to have seen to it that'" B.
I admit that this 'paradox' does not refute Professor Chisholm's
suggestion as hopelessly as the 'paradox' which he pointed out refutes
the suggestion that '0 ( ...... A --+ B)' could express a contrary-to-duty im-
perative. But it certainly speaks strongly against his suggestion.
Suppose a law-giver laid down a primary duty for his subjects to see
to it that A and a contrary-to-duty imperative to see to it that B in case
they neglect their primary duty. In order to make his intentions crystal
clear the law-giver, moreover, denies that there is a contrary-to-duty
imperative to see to it that '" B in case the primary duty were neglected. 6
What is the formally correct way of expressing this denial? One would
think that, if' ...... A --+ 0 ...... B' were a correct way of expressing the second
A NEW SYSTEM OF DEONTIe LOGIC 109
contrary-to-duty imperative, then',.., ( ,.., A -+ 0 ,.., B)' would be a correct

way of denying it. But, that ,.., ( ,.., A -+ 0 ,.., B) entails that ,.., A. Thus
from denying that a certain thing is a contrary-to-duty imperative for the
subjects it would follow logically that the subjects will neglect their
(primary) duty. This is absurd.
Chisholm shows 7 that the use of 'mixed' conditionals such as ',.., A
-+ OB' for expressing contrary-to-duty imperatives gives rise to contra-
dictions in some of the systems which allow mixed expressions. (He
mentions the systems of Mally, Prior, and Anderson.) But the observa-
tions which Chisholm makes in this context seem to me to point rather
to insufficiencies in his own suggestion concerning the form of contrary-
to-duty imperatives than to insufficiencies in those other systems of
deontic logic.
The difficulties which I have mentioned in this section constitute, in
my opinion, serious obstacles to any attempt to formalize hypothetical
imperatives (norms) by means of material implication joining a 'factual'
antecedent to a 'normative' consequent.
VI
I shall now present what I propose to call a New System of Deontic

Logic.
The vocabulary of the New System is the same as that of the Old
System save for one addition. This additional symbol is a slanted stroke '/'.
The well-formed formulae of the system are also called O-expressions.
By an atomic O-expression we now underc;tand the letter '0' followed,
within brackets, by two variables, or compounds of variables and truth-
connectives, separated by'/,. By O-expressions generally we understand
such atomic O-expressions and their molecular compounds.
An example of an atomic O-expression (of the New System) would be
'O(A/B)'. It can be read: one ought to see to it that A when B. Instead of
'when B' we can also read 'if it is the case that B' or 'should it be the
case that B'. If, for example, 'A' describes the state of affairs that the
window is closed and 'B' the contingency that it starts raining, then
'0 (A/B)' says that one ought to see to it that the window is closed should
it start raining.
The variables, or compounds of variables and truth-connectives, to the
left and right of 'f' are schemas for descriptions of two ..,ossible generic
states of affairs. The description to the left tells us, how the world ought
to be, when it is as the description to the right says that it is.
Sometimes the world is as it ought to be. It is thoroughly meaningful
to make it a duty that o (A/A). The duty to see to it that A when it is the
case that A requires us to take heed that the state of affairs in question
does not disappear.
Not always, however, is the world as it ought to be. Then the duty may
be that 0 (A/ '" A). The duty to see to it that A when this is not the case
requires us to take care that the state of affairs in question comes to be.
It should be clear from these considerations that there is no objection,
from the point of view of interpreting the formulae, to the possibility
that the same variable appears both to the left and to the right of the 'f'
in the same atomic O-expression.
The New System has three axioms:
Bl '" [0 (A/B) & O( '" A/B)].
B2 O(A & B/C) - o (A/C) & o (B/C).
B3 O(A/B v C) - o (A/B) & O(A/C).
The follow~ng example should convince us of the intuitive plausibility
of the third axiom: Suppose we are given the order to see to it that the
window is closed, should it start raining or start to thunder. Obviously
this is equivalent to being given the order to see to it that the window is
closed, should it start raining, and see to it that the window is closed,
should it start to thunder. It is an observation of some interest about the
functioning of ordinary language that the 'or' in the undistributed con-
ditioning clause of the order should become an 'and' when the clause is
being distributed.
The rules of inference of the New System are the same as the rules of
inference of the Old System (RI-R4).
VII
Consider a formula which is (an axiom or) a theorem of the Old System.
Let 'T' stand for an arbitrary tautology formed of the variables A, B, ...
and truth-connectives. Replace each atomic O-expression '0 ( -)' which
occurs in the (axiom or) theorem of the Old System by an atomic 0-

expression '0 (-/T)' where the place of ' - ' in the second O-expression
is occupied by the same variable or compound as in the first O-expression.
The formula of the New System, thus obtained, is a theorem of the New
System.
That this must be the case is easily verified. - We apply the above rule
of translation to the axioms Al and A2 of the Old System. Then we get
the formulae ...... [O(A/T)&O( '" A/T)] and O(A&B/T)+-+O(A/T)
&O(B/T). By virtue of BI and B2 and RI, these are theorems of the
New System. (Strictly speaking, they are theorem-schemas which yield
theorems for any substitution of a tautology in terms of A, B, ... for the
meta-variable T). Since the inference-rules of both systems are the same,
the translation of any formula which can be deduced from Al and A2
according to these rules must also be deducible from the translations of
Al and A2.
We can use these observations as a ground for saying that the (mona-
dic) Old System is a logic of a notion of unconditional or categorical or
absolute obligation, whereas the (dyadic) New System is a logic of a
notion of conditional or hypothetical or relative obligation. That some-
thing is unconditionally obligatory means that it is obligatory under a/l
possible circumstances (cf. axiom B3). A hypothetical obligation applies
only if certain contingencies arise.
VIII
The New System of Deontic Logic is a decidable theory.

Let there be given an O-expression (of the New System). We wish to
decide, whether it is a theorem of the system, a 'deontic tautology'.
We make a list of all the variables A, B, ... which occur in the entire
formula (O-expression). In all atomic O-expressions of the formula we
replace the variables and/or molecular compounds which occur to the
left of 'f' by their perfect conjunctive normal forms in terms of all the
variables of the list. The variables and/or components to the right of 'f'
we replace by their perfect disjunctive normal forms in terms of all the
variables. Thereupon we distribute the atomic O-expressions in accord-
ance with the rules given by B2 and B3. The atomic O-expressions, of
which the formula is a compound after the distributions, we shall call
the O-constituents of the formula (and of the original O-expression). In

each O-constituent there is, to the left of 'j', a disjunction of all the vari-
ables with or without a negation-sign in front of them. To the right there
is a conjunction (,state-description') of all the variables with or without
a negation-sign in front of them.
For the formula, thus transformed, we can construct a truth-table,
taking the O-constituents as atomic components. The distribution of
truth-values over the constituents is subject to a limitation which is im-
posed by Bl. This limitation we express in the following rule:
(i) Consider all O-constituents in which a given conjunction stands to
the right of 'I'. Form the conjunction of the disjunctions to the left of'I'
in all these O-constituents. If this conjunction is a contradiction (of
propositional logic), then, in a truth-value distribution, we cannot assign
the value 'true' to all the O-constituents under consideration.
For example: Let O[(AvB)/A&B] and O[(Av~B)/A&B] and
O[(-A v B)/A&B] and O[(-A v ~B)/A&B] be among the O-con-
stituents of a certain O-expression. (A v B)&(A v ~ B)&( ~ A v B)
& ( - A v - B) is a contradiction. Therefore we cannot assign the value
'true' to all the constituents. In every truth-distribution at least one of
them must have the value 'false'.
If the O-expression under investigation turns out to be a tautology in
a truth-table which observes the above limitation on the distribution of
truth-values, then it is a theorem of the New System of Deontic Logic.
(This assertion we shall not prove here.)
IX
Special rules must be given for the cases, when a compound to the left of
'j' in an atomic O-expression is a tautology and when a compound to the
right of 'j' is a contradiction. 8 In the first case the compound has no
perfect conjunctive and in the second case no perfect disjunctive normal
form. The normal forms 'vanish', contain no disjunction and conjunction
respectively.
We first note the following two theorems:
Tl o (AlB) -+ D(A v "" A/B).
'If under some circumstances a certain thing is obligatory, then under
those circumstances the tautology is obligatory.' The theorem immediately

follows, by the use of R3 and R4, from B2 and the tautological equi-
valence A-(A v", A)&A. It may be regarded as a limiting case of a
more general thesis to the effect that, if under some circumstances a
certain thing is obligatory, then under those circumstances also any
logically weaker thing is obligatory.
T2 o (AlB) ~ o (AlB & ,..., B).
'If a certain thing is obligatory under some circumstances, then it is
obligatory under contradictory circumstances.' The theorem follows
immediately, by R3 and R4, from B3 and the tautological equivalence
B-(B&,..., B) v B. It may be regarded as a limiting case ofa more general
thesis to the effect that if something is obligatory under some circum-
stances, then this thing is also obligatory under any logically stronger
circumstances.
By virtue of R3, we can replace A v ,..., A in T1 by any arbitrary tautol-
ogy, and B&,..., Bin T2 by any arbitrary contradiction.
If 'T' represents an arbitrary tautology of the variables A, B, ... we
can represent an arbitrary contradiction by , '" T'. Let us use'S' and' U'
to represent arbitrary synthetic formulae, i.e. non-tautologous and non-
contradictory expressions composed of variables A, B, ... and truth-
connectives.
Assume now that an O-expression, which we wish to test for theorem-
hood, contains atomic O-expressions of the form '0 (TIS)' or '0 (S/,..., T)'
or '0 ( T/ ,..., T)'. 9 We replace the expressions to the left of 'I' in atomic 0-
expressions of the type '0 (S/,..., T)' by their perfect conjunctive, and the
expressions to the right of 'I' in atomic O-expressions of the type '0 (T/S)'
by their perfect disjunctive normal forms in terms of all the variables in
the entire (compound) O-expression under investigation. Thereupon we
distribute the O-expressions in accordance with B2 and B3. The atomic
O-expressions which we get after these distributions shall count as 0-
constituents of the O-expression under investigation. (All the other
constituents are obtained by distributing atomic O-expressions of the
form '0 (S/U)'.)
The additional restrictions on the distribution of truth-values over
constituents are as follows:
(ii) If some constituent of the form '0 (S/U)' is assigned the value 'true'
then all constituents, if there are any, of the form '0 (T/ U)' and of the
form '0 (S/", T)' must be assigned the value 'true'.
(iii) If some constituent of the form' 0 (T/ S)' or '0 (S/ '" T)' is assigned
the value 'true', then all constituents, if there are any, of the form
'0 (TI '" T)' must be assigned the value 'true'.
Consider the two imperatives expressed by 'O(AIT)' and '0 (BI'" A)'.
The first is a categorical or unconditional order under any circumstances
to see to it that A. The second is a hypothetical or conditional order to
see to it that B, should it be the case that '" A.
It is easily shown that the categorical imperative does not entail the
hypothetical one, but that the two are logically independent. The cate-
gorical order can be expanded as follows: O[(A v B)&(A v'" B)/A &B
v A &'" B v '" A &B v '" A &,.., BJ. The hypothetical order again is
equivalent to 0 [(A v B) & ( '" A v B)/'" A &B v '" A & '" BJ. The distri-
bution-axioms B2 and B3 may be used for splitting up the two expressions
into O-constituents. Then the categorical order becomes a conjunction
of 8 and the hypothetical order a conjunction of 4 O-constituents. Two
of the constituents of the hypothetical order are not among the consti-
tuents of the categorical order. This suffices to show that the two orders
are logically independent.
Shall we say that '0 (B/,.., A)' expresses the contrary-to-duty impera-
tive to see to it that B in case a duty-bound subject has neglected his
primary duty to see to it that A? We shall not say this. For, there is
nothing in the form of the imperative to show that the state of affairs
that '" A has come about as a result of neglect of duty. To show this we
should have to make use of a symbolism which contains names of agents
and has means of distinguishing between various ways in which states
of affairs may come to be. Such a symbolism can be developed, - but we
shall not try to develop it here.
Contrary-to-duty imperatives are a special class of hypothetical im-
peratives. The logical structure of hypothetical imperatives, I shall
maintain, is that of a dyadic obligation-operator which obeys the axioms
BI-B3 of the calculus here called the New System of Deontic Logic.
Neither the formula '0 ( '" A -. B)' of the Old System, nor the 'mixed'
formula • "-' A -+ OB' proposed by Professor Chisholm captures the logical

form of the hypothetical norm that one ought to see to it that B, should
it be the case that "-'A.10 The formula '0 (B/"-' A)' of the New System
captures it. If the reason, why it is not the case that A, is that the agent
to whom the hypothetical norm applies has neglected a duty of his to
see to it that A, then the hypothetical norm is, for him, a contrary-to-
duty imperative.
• * *
XI
The immediate purpose which I had in mind when constructing the New
System was to deal with certain difficulties which had been pointed out
by Professor Chisholm. l l The difficulties in question centered round a
notion for which Chisholm had coined the name "contrary-to-duty im-
peratives". A contrary-to-duty imperative, generally speaking, tells us
what ought to be done when a certain (primary) duty has been neglected.
As shown by Chisholm, certain systems of deontic logic, among them
my Old System, cannot adequately formalize such imperatives. In my
paper I wanted to show that the New System captures the general form
of contrary-to-duty imperatives.
What was said on contrary-to-duty imperatives in my paper remains,
for all I can see, untouched by the objection to the New System which I
shall presently mention and by the amendment of the system which I am
going to propose.
XII
In the New System one can prove certain things which are counter-
intuitive and therefore not wanted in a sound deontic logic. I am much
obliged to Mr. P. Geach for his having brought this to my attention.
Let there be a norm to the effect that one ought to see to it that A is
the case, should it be the case that B. In symbols: O(A/B). By virtue of
the Rule of Extensionality, o (A/B) is equivalent with O(A/B&C v B&
'" C). This, by virtue of B3, is equivalent with O(A/B&C)&O(A/B&
'" C) which entails o (A/B& C). Thus we have proved the formula
o (A/B)-+ O(A/B&C).
By similar arguments we prove 0(", AIC)-+ O( '" AlB & C). This is
equivalent with '" 0 ( '" AlB & C) -+ '" 0 ( '" A/C).
By virtue of Bl, we have", [O(A/B&C)&O( '" A/B&C)] which is
equivalent with O(AIB&C)-+ '" O( '" A/B&C).
O(AIB)-+ O(A/B&C) and O(A/B&C)-+ '" O( '" A/B&C) yield to-
gether O(A/B)-+ ",O( '" A/B&C).
O(A/B)-+ '" O( '" A/B&C) and", O( '" A/B&C)-+ '" O( '" A/C) yield
together 0 (A/B) -+ '" 0 C·v A/C).
Herewith it has been proved that, if there is a duty to see to it that A
under circumstances B, then there is no duty to see to it that not-A under
circumstances C. For example: It has been proved that, if there is a duty
to see to it that a certain window is closed should it start raining, then
there cannot be a duty to see to it that the window is open should the
sun be shining. This is manifestly absurd. Generally speaking: From a
duty to see to a certain thing under certain circumstances nothing can
follow logically concerning a duty or not-duty under entirely different,
logically unrelated, circumstances. Least of all should one be able to
prove that there is under those unrelated circumstances a duty of contra-
dictory content.
XIII
Can the root of the trouble be traced and the calculus put right? I think
this is possible.
The source of the trouble, as I see it, resides in axiom Bl. This axiom
was meant to be an analogue of the axiom
Al '" (OA & 0 '" A)
of the Old System of Deontic Logic, i.e. of the monadic calculus of un-
conditional or categorical or absolute norms (duties).
The analogue, however, was falsely conceived.
An expression formed of the letter '0' in front of a disjunction of single
variables and/or their negations will be called a (normal) monadic 0-
constituent. n variables determine 2" monadic O-constituents. We shall
also call the constituents the O-units of the deontic space determined by
n variables (for generic states of affairs).
One variable A determines a deontic space of two O-units, OA and
0", A. Axiom Al thus says that the two deontic O-units of the deontic
space determined by one variable (state) cannot both be valid. At least

one of the units must be invalid.
Two variables, A and B, determine a deontic space of 22 =4 O-units.
The units are O(A v B), O(A v - B), 0(- A v B), and O( - A v - B).
If we replace A in Al by its conjunctive normal form in terms of A and B
and apply the conjunctive distributivity of the operator 0, it immediately
follows that not all four O-units can be valid. Generally speaking: In a
given deontic space not all the O-units can be valid.
(If we accept the interdefinability of permission and obligation accord-
ing to the schema P = - 0 -, then the above principle is equivalent to
the rule that, in a given deontic space, at least one P-unit must be valid.)
The principle that, in a given deontic space, not all the O-units can be
valid is to hold unchanged also for the New System of Deontic Logic,
i.e. for the dyadic calculus of conditional or hypothetical or relative
norms (duties).
Let there be a set of n variables. Consider a disjunction of some m of
them and the negations of the remaining ones and a conjunction (state-
description) of some k of them and the negations of the remaining ones.
The letter '0' followed (within brackets) by this disjunction and conjunc-
tion (in that order) separated by 'f' will be called a (normal) dyadic 0-
constituent. We shall also call it a dyadic O-unit of the deontic space
determined by the n variables (generic states of affairs).
The number of dyadic O-constituents which is determined by n variables
is 2 ·2 or 2 2". For n=2 this number is 16. For n=l it is 4. The four
ft ft
O-units (-constituents) determined by the single variable A, are 0 (AlA),

O(AI-A), O(-AIA), and O(-AI-A).
We replace axiom Bl by an amended form Bl' which says that the
four dyadic O-units of a deontic space determined by a single variable
cannot all be (conjunctively) valid. In symbols:
Bl' - [O(AIA)&O(AI- A)&O(- AIA)&O(- AI- A)].
The validity of the principle for the general case of n variables (states)
follows immediately from B l' with the aid of the distribution-axioms B2
and B3 and the Rule of Extensionality.
By virtue ofB3, Bl' is equivalent with - [0 (AlA v", A)&O( '" AlA v
'" A)]. By virtue of the Rule of Extensionality, we can substitute 'B v '" B'
for 'A v - A' in this formula. A second application ofB3 then yields the
formula -[O(AIB)&O(-AIB)&O(AI-B)&O(-AI-B)]. It is in-

structive to compare this theorem with the original axiom Bl.
On the basis of considerations about truth-value distributions and 0-
constituents, it may be shown that neither Bl nor the objectionable
formula o (AIB)-. - O( '" AIC) is a theorem of the amended system.
If, therefore, we reject Bl and replace it by the essentially weaker axiom
BI', the counterintuitive results mentioned in Section XII are no longer
derivable.
XIV
The New System of Deontic Logic, thus amended, caters for a logical
possibility, to which the system in its original form cannot do justice.
This is the possibility of conflicting duties (obligations).
We shall say that duties conflict, if they require the doing of conjunc-
tively impossible actions under the same circumstances. For example:
a duty to see to it that A when B and a duty to see to it that '" A when B
are a pair of conflicting duties. In symbols: 0 (AI B) and 0 ( '" AlB) are
conflicting. According to the original version of the system, they cannot
both be duties. This was excluded by axiom Bl. According to this system,
a (genuine) conflict of duties was therefore a logical impossibility. This
it obviously is not. One thing which the derivation of the absurdity in
Section XII shows, is the necessity of allowing for the possibility of
conflicting duties in a sane system of conditional norms.
The axiom Al of the Old System excludes conflicting duties too. Does
it mean that the system is not sound? It does not mean this. For, it may
truly be regarded as a logical impossibility that absolute (categorical, un-
conditional) duties or obligations should conflict. (Unconditional duties
under different laws or systems of norm may, of course, conflict, in the
sense that they impose logically contradictory or contrary demands on
an agent. Such cases, however, are logically uninteresting and should
better not be regarded as genuine 'conflicts of duties' at all.)
It is logically absurd that a man should have conflicting duties whatever
his circumstances are. Conflicting duties arise only under special circum-
&tances. Such circumstances constitute what may be called a (moral)
predicament.
Of particular interest both from a logical and from an ethical point of
view is the predicament which arises when a man promises to do the
forbidden. (Jephtha's case in the Book of Judges.) He who has promised

to do something which is forbidden, ought, by virtue of his promise, to
do the very thing which, by virtue of the prohibition, he ought to abstain
from doing. The predicament, however, is not there simpliciter, 'in itself',
independently of circumstances. The predicament arises thanks to the
doing of the agent, i.e. thanks to his giving a certain promise.
It may be shown that if the act of an agent gives rise to conflicting
duties, then this act is itself something from which the agent has a duty
to abstain. (Promising the forbidden is itself forbidden.) The proof will
not be given here. 12
Academy of Finland
NOTES
• Sections I-X of this paper are a reprint of 'A New System of Deontic Logic',
Danish Yearbook 0/ Philosophy 1 (1964) 173-182; Sections XI-XIV are a reprint of
sections 2-5 of 'A Correction to a New System of Deontic Logic', Danish Yearbook
o/Philosophy 2 (1965) 103-107. Reprinted by permission of the author and the publisher.
1 Roderick M. Chisholm, 'Contrary-to-duty Imperatives and Deontic Logic', Analysis
24 (1963) 33-36.
2 G. H. von Wright, 'Deontic Logic', Mind 60 (1951) 1-15.
S There is a (fragmentary) discussion of these notions in my book Norm and Action,
Routledge and Kegan Paul, London, 1963, pp. 22-27.
4 Chisholm, ibid., p. 34: "It is clear that we must use instead a conditional with an
obligatory consequent and tell him 'If you steal then it is obligatory that you return
the money'."
5 I say "seems to entail" because I am not quite certain whether 'A' and '"" A' can
be correctly interpreted as saying that the agent has fulfilled, respectively neglected,
his duty to see to it that A. If they cannot be thus interpreted, however, then another
objection to Chisholm's suggestion seems to be forthcoming. This new objection is
that' "" A -+ 0 B' does not show that the state of affairs described by '"" A' is supposed
to have come about as a result of neglect of duty - and therefore cannot express a
contrary-to-duty imperative. (Cf. below Section X.)
6 Chisholm, ibid., p. 34: "We may also wish to tell him, if the need arises, 'It is not
true that, if you steal then it is obligatory that you steal again'."
7 Ibid., pp. 34-35.
8 I am indebted to Dr. Lars Svenonius for his having drawn my attention to a defect
in an earlier attempt to formulate these rules.
9 Because of R3 we may regard the contradiction to the right of 'j' as the negation of
the very tautology to the left of 'j' .
10 The view that ',,4 -+ OB' is the logical form of a hypothetical norm has been inter-
estingly argued by Professor Erik Stenius in his paper 'The Principles of a Logic of
Normative Systems' in Proceedings 0/ a Colloquium on Modal and Many- Valued Logics,
Acta Philosophica Fennica 16 (1963), esp. pp. 256-260. It seems to me, however, that
Stenius's view is not capable of doing full justice to the logical peculiarities of this type
of norm.
11 Roderick M. Chisholm, 'Contrary-to-duty Imperatives and Deontic Logic', Analysis
24 (1963) 33-36.
12 For the proof of an analogous principle in the monadic calculus see my paper
'Deontic Logic', Mind 60 (1951), p. 13.
BENGT HANSSON
AN ANALYSIS OF SOME DEONTIC LOGICS·
I. INTRODUCTION
The purpose of this article is to investigate the structure and to discuss

the applications of deontic logics of von Wright's type as presented in
[15] and [19]. The 'paradoxes' which arise in these logics seem to indicate
that the axioms reflect only some special senses of the words 'obligation'
and 'permission'. I believe that the questions of what these special senses
are can be clarified by exhibiting a simple picture of the formal structure
of these logics.
Deontic logics may also be constructed in a dyadic form where the
concept of obligation is made relative to some circumstances. Such
systems have been proposed by von Wright (in [16] and [17]) and Nicholas
Rescher (first in [13]). I will prove that there is essential disagreement be-
tween von Wright and Rescher and I will propose a third dyadic system
which, in my opinion, best preserves the main ideas of von Wright's
monadic version. In doing this, however, I do not claim that this dyadic
deontic logic is a good one in any general sense. It is outside the scope
of this article to discuss or propose alternatives to von Wright-type
deontic logics and I do not want to commit myself to any specific view
on their merits and demerits. But, as I hope, this article would be relevant
to such a discussion, because one of the things I try to do is to point out
what von Wright-type logics really are about.
II. THE LANGUAGE
I assume the existence of a basis logic (henceforth BL) which may be the
propositional calculus or the first-order predicate calculus with or without
constants, or some related system. It is essential that valuations and
validity can be defined in the usual way and that BL is complete in the
sense that every valid formula is a theorem. I also assume that it is well-
known what constitutes a well-formed formula and a theorem of BL
R. Hllpinen (ed.). Deontic Logic: Introductory and Systematic Readings. 121-147. All rights reserved.
Copyright © 1970 by D. Reidel Publishing Company. Dordrecht-Holland.
122 BENGT HANSSON
respectively. I will use/, g, h etc. as meta-variables rang:ng over the set

of formulas of BL.
The deontic logic under study will be called the standard deontic logic
(SDL) based on BL. The set of formulas of SDL is the smallest set ful-
filling the following requirements: (i) whenever/is a formula ofBL, then
0/ is a formula of SDL; (ii) the negation of any formula of SDL is a
formula of SDL; (iii) the disjunction of any two formulas of SDL is a
formula of SDL.
The intended reading of the operator '0' is 'it is obligatory that'. Two
other operators, F ('it is forbidden that') and P ('it is permitted that') may
be defined as follows: F = def 0 ....., ; P = def ....., F. Truthfunctional connectives
between formulas of SDL may be defined as usual·.
This language is poorer than that used in most deontic logics: (a)
iterated operators are not allowed; (b) there are no mixed formulas, e.g.
of the type 'p v Op' (c) even if BL is the predicate calculus only truth-
functional connectives are used between formulas of SDL, i.e. no formula
of SDL ever occurs within the scope of a quantifier. But this is intentional,
because then SDL can be regarded as a subsystem of most other systems.
Besides, almost all philosophical problems discussed in connection with
deontic logic are expressible in this language.
III. THE STATUS OF FORMULAS
The formal structure of deontic logic can of course be investigated with-

out any discussion of what kind of entities the formulas stand for. But
this becomes important when it comes to the question of applications.
There are two kinds of formulas involved, viz. those of BL and those of
SDL, and therefore two interpretation problems.
Formulas of BL are normally thought of as propositions, entities being
true or false. But we think of the property of being obligatory as a
property of acts and acts are not propositions. Either we have to re-
interpret BL or to say that 'obligatory' is a property of descriptions of
acts. Both solutions seem a little clumsy but none more than the other.
I will use the second one on the sole ground that it seems to be common
usage. Descriptions in general will be discussed again below in Section VI.
Formulas of SDL are in a way treated as if they were propositions,
because truth-functional connectives are used to form compound for-
AN ANALYSIS OF SOME DEONTIC LOGICS 123
mulas. But a phrase like 'it is obligatory that p' is generally not considered
to be a true or false statement. On the other hand a deontic statement is
not simply an imperative (although some philosophers hold the view that
there is a sense in which it is equivalent to one), because one can point
out to a person that he ought to do so-and-so without actually telling
him to do it. I will here take the view that deontic statements (formulas
of SDL) are descriptive, that they describe what is obligatory, forbidden
and permitted respectively, according to some (undetermined) system of
norms or moral or legal theory. The proper reading of e.g. 'P' should
then be: 'Norm-system X permits that .. .'. The deontic axioms which
will be discussed later, then, do not have the status of logical truths, but
they express properties of the norm-system used. Those who are attracted
by the axioms may then, if they so want, regard them as criteria of ratio-
nality or of inner coherence of norm-systems or moral or legal theories.
This descriptive interpretation has one advantage in addition to
making formulas of SDL into propositions: it makes clear that deontic
logic is a tool of meta-ethics and not a part of ethics proper.
The interpretation could be supplemented with a reference to the
person or class of persons to which the norms are directed. If this is done,
I will presuppose that the reference is held constant throughout one and
the same context.
In giving this interpretation I do not claim that it is the only one which
is good or interesting, only that it provides a ground for discussing the
philosophical implications of certain results about the structure of von
Wright-type deontic logic. Discussions based on other interpretations
may yield interesting results and I do not '.'{ish to exclude them.
IV. LOGIC BY WAY OF SETS
A valuation of BL is a function which assigns a truth value to each

formula of BL in such a way that every consequence of a formula which
has the value true also has the value true and such that a formula and
its negation always have different values. If one attaches a meaning to
each symbol of BL, a valuation may be regarded as a possible world
(possible in the sense of logically possible). Since this article is primarily
concerned with interpretations of deontic logic, I will use this term
because I think its vividness will more than compensate for the risk of
124 BENGT HANSSON
confusion. But the reader should be aware that the use oftJ-.e term involves
no ontological commitments.
If/is any formula of BL, let T(f) be the set of possible worlds in which
I is given the value truth. Let t be the set of all possible worlds. "'" I is
given the value true in exactly those possible worlds where I is not true,
i.e., T(-/)=t-T(f). It is also obvious that T(p&q)=T(p)rtT(q)
and T(p v q)=T(p)uT(q).
IfI is a theorem of BL we know that T(f) = t. And if T(/) = t we know
that I is a theorem because BL is complete. The completeness theorem
for BL also says that I and g are provably equivalent if and only if
T(f)=T(g). So if one is just concerned about the content of a formula
and not about its form one could just as well talk about the set T(f)
instead off I will do so because one can always replace a formula by a
provable equivalent in von Wright-type deontic logics. I will further drop
the function symbol T, so that J, g etc. will be treated directly as sets.
Since negation corresponds to complement, conjunction to intersection
and disjunction to union, it does not matter which set of concepts one
uses. I will use t as short for theorem (a tautology) and k as short for
any formula which is the negation of a theorem (a contradiction). k then
is simply the empty set.
If I and g are formulas, then 1--+ g is provable in BL if and only if
I r;;;. g. Note that the lormula 1--+ g is a set (viz. "'" lug), but the fact
that I --+ g is provable is the fact that I is a subset of g.
V. THEORIES AND THEORY-SETS
Every formula is a subset of t, but not every subset of t is a formula.

I will need a more general kind of sets than formulas, sets called theory-
sets, which are closely related to theories. By a theory I mean any non-
empty set T of formulas fulfilling the following conditions:
(i) if I belongs to T and if I r;;;. g and if g is a formula, then g
belongs to T
(ii) ifI and g belong to T, then I rt g belongs to T
A theory T is inconsistent if it is the set of all formulas; otherwise it is
consistent. Tis axiomatized by S if S is a set of formulas such that there
exists, for every I in T, a finite number of formulas in S, 11> ". /", such
that the intersection fl n ... n f .. is a subset off, and if this is impossible

for f's outside T. Every theory can be axiomatized, e.g. by taking S as T,
so the concept ofaxiomatization is interesting only insofar as it is possible
to find an S with some interesting additional property, e.g. finitude.
A set BT is a basis for the theory T if and only if T consists of exactly
those formulas f, for which BT S;; f is true.
THEOREM 1: Every theory has a basis.
Proof· Define BT as the intersection of all formulas in T. It is imme-
diately clear that BT s;; fholds for every fin T. Now assume that formula
fhas BT as a subset in order to prove thatfbelongs to T. I will use the
compactness theorem for BL which says: if M is a set of formulas such
that every finite intersection of members of M is non-empty, then the
intersection of all members of M is non-empty. Define T' as Tu {- f}.
The intersection of all members of T' is BT n - f =k. Therefore there
exists a finite number of sets in T', the intersection of which is empty.
Case 1: - f is not among these sets. Then some finite intersection of
sets in T is empty. k then belongs to T by clause (ii) in the definition of a
theory. By (i) then every formula belongs to T, among themf
Case 2: - fis among these sets. Let the sets befl' ... '/'" - f;/t being
in T.fln ... n/"n-f=k means fln ... n/" s;;f By (i) and (ii)fmust
be in T.
COROLLARY: A theory is consistent if and only if it has a non-empty
basis.
There may exist several bases for a certain theory. But the one which
is constructed as in the proof of Theorem 1 plays a certain role, because
it is the largest one. For let B be any other basis for the same theory T.
B s;; f for every f in T implies that B is included in the intersection of all
suchf's, i.e. in BT. I will call every set which, like BT, is the intersection
of a class of formulas of a theory a theory-set.
THEOREM 2: Among the bases for a certain theory T there is exactly
one which is a theory-set.
Proof· The existence is already proved. Now assume that B is another
basis for T which is a theory-set. Then B is a proper subset of BT. Let
M be the class of formulas the intersection of which is B. Case I: Mis
a subset of T. But then BT is the intersection of all members of M and
perhaps some more formulas. Then BT is a subset of B, which is a contra-
diction. Case II: M is not a subset of T. Then there exists some formula,
126 BENGT HANSSON
say f, which is in M but not in T. But B ~ fimplies that/is in T because

B is a basis for T. Contradiction again.
The theory-set which is a basis for a theory will be called the canonical
basis. All bases in the following will be canonical except when the
contrary is explicitly stated.
The following two lemmata will be used in proofs in later sections.
LEMMA 1: If a is a theory-set distinct from t, then there exists a formula
f, such that a ~ f c t.
Proo!' a is the intersection of formulas, at least one of which is not t
(otherwise a would be t). Any such formula will do asf
LEMMA 2: If x and yare two distinct elements of t, then there is a
formulaf, such that x belongs to f and y belongs to '" f
Proof: Since x and yare distinct valuations of BL, there is at least one
formula, say g, which has different truth values according to x and y
respectively. If g is true in x, then g will do as f, otherwise '" g will do.
VI. DESCRIPTIONS
In asserting a formula one really says that the real world is one of the
possible worlds which are elements of the formula, or, equivalently, that
some possibl~ worlds (viz. those in the complement of the formula) are
ruled out as not being actual. We may say that a formula is a description
of the world, but only a partial one, because we can never identify a
possible world as the real one only on the evidence of a single formula.
The information carried by a formula is simply that the range of possible
worlds among which we have to look for the real one is narrowed down
a bit. A complete description would be one which tells us exactly which
set the real world is, i.e. a set containing only one element. But no one-
element set is a formula. What other kind of descriptions do we have
than those provided by formulas? Sometimes we not only assert a single
formula, but a whole class of formulas at the same time. Such a descrip-
tion is very like a theory, because if asked we would probably also assert
finite conjunctions and logical consequences of the formulas in the class.
Therefore such a description can be identified with an assertion of the
basis of the theory generated. This is a slight extension of the use of
'assert'; to assert a set will mean to assert that the real world is one of
the possible worlds in that set.
Among the descriptions of this new kind we find the complete de-
scriptions, the singleton sets, because every such set is a theory-set (as
can easily be proved).
But we can think of still more general descriptions. We can, e.g., assert
that at least one of the formulas in a certain class is true. This assertion
can be represented by the union of the formulas and this set need neither
be a formula nor a theory-set. And we may have an infinite class of
theories and assert that at least one of them is true (i.e., that in a certain
set of classes of formulas, there is at least one in which every formula is
true), this assertion being represented by the union of the bases, which
again may be a new kind of set. And we may go on and claim that every
Borel set over the set of formulas is a description.
All these kinds of descriptions have only one thing in common: they
are (in a more or less complicated way) generated from the set of for-
mulas, and therefore expressible (admitted in a very general sense) in the
metalanguage of BL.
But the most general kind of description is an arbitrary subset of t.
It provides the information that possible worlds outside that set are in
fact not actual. But for these descriptions we do not in general have any
connection with the metalanguage of BL, so they are in a certain sense
inexpressible. Since the cardinality of the class of formulas is strictly less
than that of the class of theory-sets, which in turn is less than that of all
subsets of t, both formulas and theory-sets are indeed scarce among the
subsets of t. So there is a lot to be said which cannot be said in formulas
or theories. But in one respect theory-sets are nevertheless sufficient, viz.
as regards the class of formulas one is prepared to assert given a certain
description of the world. Such a class should have the structure of a
theory and is therefore fully determined by its basis, a theory-set.
VII. AXIOMS FOR SDL AND DEONTIC BASES
Now when the structure of BL has been discussed at some length, the
structure of SDL will be revealed quite easily. The following two axioms
are suggested by von Wright in [15] and [19]:
Al O(pnq)++Op&Oq
A2 '" O(k)
128 BENGT HANSSON
Sometimes the following axioms are used:

Bl Op -. "'" 0 "'" p (i.e. Op -+ Pp)
B2 O(p-+q)&Op-+Oq
Every formula which is obtained from a theorem of propositional logic
by substituting formulas of SDL for all propositional variables, is also
an axiom of SDL. We accept any normal set of rules of inference which
includes substitution and replacement. The axiom systems AI-A2 and
BI-B2 are not equivalent, but both Bl and B2 are provable in AI-A2,
while neither Al nor A2 is provable in BI-B2. Nevertheless the structures
of the two systems are very similar, and they become equivalent if the
following axiom is added to both of them:
AO = BO O(t)
Sometimes this axiom is added to the B-system as an inference rule
instead: iffis provable in BL, then Of is provable in SDL. The content
of AO and BO is very small; the only model of AI-A2 in which AO is not
true is a trivial one where Of is false for every f BI-B2 without BO
admits a few more models, but none of them is compatible with the
intended interpretation. Even if von Wright did not propose AO, it seems
fair to define SDL as the logic which has AO-A2 as axioms and the
language described in Section II, and still claim that SDL is essentially
what von Wright meant.
THEOREM 3: Iff S;; g, then Of -+ Og is provable in SDL.
Proof' Sincefequalsf rig, Of implies Of &Og by AI.
THEOREM 4: In every model of SDL, the set off's such that Of is true
forms a consistent theory.
Proof' Al proves property (ii), Theorem 3 proves property (i). and A2
proves the consistency.
COROLLARY: Every model of SDL is completely described by a non-
empty basis.
I will discuss the implications of this corollary in the next few sections.
VIII. KANGER'S AND ANDERSON'S MODELS
In [11] Stig Kanger discusses topics related to deontic logic. He uses a

propositional constant Q with the intended interpretation of a description
of 'what morality prescribes'. He then mentions the following definition

of the obligation operator:
Of = defN(Q f) -+
where 'N' is the modal necessity operator. There are no difficulties in

principle to extend SDL to include modal operators, but in order to
avoid notational complications I will interpret necessity as provability in
BL. The definition then reads:
Of = defQ £; f
Kanger then proves that '0' thus defined fulfils the axioms of SOL,
provided Q is not k. So every model constructed along these lines is a
model of SDL.
Alan Ross Anderson presents a similar idea in [1], [2], [3] and [6].
He uses a constant S which expresses some 'sanction' or 'wrong thing' and
defines something as obligatory when the absence of it implies S:
Of = def"" f entails (in some sense) S
The kind of entailment is different in different papers; the problem is
surveyed in [6]. If we take entailment in the sense of provable implication,
Anderson's definition becomes:
Of= def"" f £; S
which is equivalent to
Of = def"" S £; f
So if we identify S with ,.., Q we have Kanger's definition again.
What about the possible worlds which are elements of Q? Since Q is
'what morality prescribes' the possible worlds in Q must be (morally)
superior. There may be many worlds in Q, but the differences between
them cannot be morally relevant, or either Q would have excluded some
of them. I will label the possible worlds Q (morally) ideal worlds. Thus
Kanger and Anderson have proved that if we dub some possible worlds
ideal and say that something is obligatory whenever it is true in all ideal
worlds, then this will be a model of SOL. Now the corollary to Theorem
4 in Section VII proves the converse result: every model of SOL must
have this structure. The talk about ideal worlds is not only an illustration,
but the illustration of what SDL is. If we accept von Wright type deontic
logic, we have a deontic logic which can be described by ideal worlds.
130 BENGT HANSSON
Now, what does 'P' mean in terms of ideal worlds? Plmeans '" 0 "'" f,
i.e. "'" lis not obligatory, i.e. the basis is not a subset of "'" f, i.e. the basis
and I have a non-empty intersection. Something is permitted, then, if
and only if it does not exclude all ideal worlds. 'Forbidden', accordingly,
is to be interpreted 'does exclude every ideal world'.
Permissions in the sense of SOL must be distinguished from rights in
the sense of claims. Permission to do I means in SOL 'if you do f, you
will not be blamed; you have not broken any rules'. A right normally
entails not only permission, but also obligations for other persons to
refrain from bringing about,..., f. But in SOL it is quite possible that some
person is permitted to do I and another to do '" J, and he who had the
permission to do I must not complain if somebody else did ,..., I thus
making it impossible for him to use his permission. P is then a weak kind
of permission in SOL, but the definition of P in terms of 0 is not essential
to the system. The axioms are formulated in 0 and one may regard SOL
as a system exclusively about obligations if one is not satisfied with the
weak senses of permission.
The case when the basis is a one-element set deserves some special
attention. Then 'having the basis as a subset' is tantamount to 'having
non-empty intersection with the basis' and therefore everything permitted
is also obligatory and vice versa. The norm system or moral or legal
theory which has such a basis leaves no room for free choice without
breaking the rules; for every f, eitherf or ,.., fis obligatory. This situation
arises when the norm system is too detailed: since every change in the
ideal world makes it non-ideal, every formula expresses something
morally relevant. But if we assign meaning to the formulas in such a way
as to cover all aspects of the world, then some formula will mean 'Jones
buys a green pencil' and some other 'Jones buys a yellow pencil'. If e.g.
the first one is true and the second one false in the ideal world, there seems
to be no moral reason why it should cease to be ideal if the truth values
were reversed. So norm systems with one-element bases may be looked
upon with some suspicion.
IX. A PARADOX
Some theorems of SDL have been called paradoxes. This means of course
that they seem counterintuitive, although they are derived from intuitively
acceptable axioms. The general line of a 'solution' is then to point out

that the concepts involved are ambiguous; one sense is employed when
the axioms are judged acceptable and another when the theorem is said
to be counterintuitive. Since we have now a fairly good idea of the
structure of SDL, we also know what its axioms commits us to. I will
now discuss some of these paradoxes and how they should be interpreted
if one uses the concepts of 'obligation' and 'permission' in the sense of
SDL.
I think that there are two important aspects of the paradoxes. First,
they seem to make too much obligatory; second, certain obligations
cannot be symbolized in SDL. The paradox named after Alf Ross
illustrates the first point. The theorem thought paradoxical is:
Of -+ 0(/ U g)
Why is this paradoxical? Let/stand for 'you help Mr. A' and 9 for
'you kill Mr. A'. If B tells C 'it is obliptory that you help Mr. A' the
theorem seems to justify C if he deliberates as follows: what B told me
implies 'it is obligatory that I help Mr. A or that I kill Mr. A'. Now this
obligation can be fulfilled by killing Mr. A. Therefore I will kill Mr. A
and then tell B that I fulfilled the obligation.
An immediate objection is: C has fulfilled the obligation '0 (J u g)'
but not the obligation B told him about, i.e. 'OJ'. Furthermore C has
performed g, but (as we hope) 9 is forbidden, which, in our deontic logic,
is quite compatible with O(/ug). C is then wrong because (a) he has
not fulfilled the obligation which was pointed out to him, but another
one; (b) he fulfilled this obligation in such a way that he at the same time
performed a forbidden act. So just fulfilling an obligation is not enough;
but it must be fulfilled subject to the constraint that no forbidden act
be done at the same time.
But, admitted that C is wrong, we still have to face the problem: do
we really want to assert 'it is obligatory that you help Mr. A or that you
kill Mr. A'? We could answer that we have to, because we have agreed
to call something obligatory if it is true in all ideal worlds and the de-
scription of the disjunctive act we are discussing is so. But this answer
does not explain the queer feeling that arises from the obligation. Perhaps
the following observation might be of some help: What we often call
descriptions of acts are not descriptions of how the act is performed but
132 BENGT HANSSON
of the result of the act. It is in general possible to proceed in several different

ways to achieve the same goal and it is certainly so if the result is described
as a disjunction. That somebody asserts an obligation does not mean
that he approves of every way of making the obligatory formula true.
Specifically, to fulfil the obligation to help or kill Mr. A by killing Mr. A
would be an unacceptable way, but doing it by helping him is all right.
In fact 0 (f u g) is then not more paradoxical than OJ, because even this
obligation can be fulfilled in unacceptable ways, e.g. by helping the
crippled Mr. A downstairs by kicking him. Let h stand for 'you kick
Mr. A down the stairs'. Already the fact that Of is equivalent to 0((1 nh)
u (I n "'" h)) constitutes Ross' paradox, which then is not specific to von
Wright type deontic logics.
A possible objection to this solution may run as follows: thus inter-
preted deontic logic is of no worth, because we cannot tell whether an
act is good or not even if we know that its result is obligatory. In other
words: the logic seems to fail to point out exactly those ways to fulfil a
certain obligation which are all right. But we must look not only on the
obligations uttered or asserted, but on the deontic system as a whole.
That killing Mr. A is a wrong way of 'helping or killing' him is revealed
by the fact that he then necessarily fails to perform another obligatory
act, ...., g, to refrain from killing Mr. A. So even if the good ways of
fulfilling a certain obligation are not determined by that very obligation,
they are nevertheless determined by the set of all obligations.
X. ANOTHER PARADOX AND DYADIC SYSTEMS
Ross' paradox tried to say that too many acts were obligatory. Other
paradoxes arise because we do not seem to be able to express certain
obligations in SDL. Among them is the paradox of contrary-to-duty
obligations, discussed by Roderick M. Chisholm in [10]. I will closely
follow Lennart Aqvist's presentation of it in [7].
Consider the following four sentences:
I It ought to be that Smith refrains from robbing Jones.
II Smith robs Jones.
III If Smith robs Jones, he ought to be punished for robbery.
IV It ought to be that if Smith refrains from robbing Jones he
is not punished for robbery.
AN ANALYSIS OF SOMB DBONTIC LOGICS 133
A way to formalize these sentences is the following, where we tenta-

tivelyextend SOL to include mixed formulas:
(1) 0 '" f
(2) f
(3) f -+ Og
(4) O( '" f -+ '" g)
But from the second and third formula we may deduce Og and from
the first and fourth 0", g, which is a contradiction in SOL, although
I-IV are perfectly consistent intuitively. So this is not a good formalization.
Try this way instead: III and IV are in a sense parallel; it is mere
coincidence that 'ought' appears in the middle of III but at the beginning
of IV. III and IV should then be formalized in similar ways. Either (3)
should be replaced by (3'): 0(1 -+ g), or (4) by (4'): '" f -+ 0", g. But
(3') is a provable consequence of (1) although III is not a consequence
ofI; similarly for (4') and (2). So these ways are not good either.
The obligation in III is of a different type compared to the one in I.
It expresses a secondary duty which applies when the primary duty in I
has been violated. It seems to be difficult to express such a secondary
duty in SOL. This fact suggests the use of a new primitive dyadic operator
O( -/-), where o (f/g) is to mean 'fis obligatory in circumstances g'.
Systems of this type have been proposed by von Wright in [17] and Rescher
in [13].
In the remaining part of this article I will discuss dyadic systems. This
does not mean that I commit myself to the view that it is impossible to
express secondary duties in SOL (l have no definite opinion on that
matter), but if it is possible then it is a criterion of adequacy that the
acceptable axioms for 0 ( - / - ) be theorems.
XI. RBSCHBR'S AXIOMS AND THB STRUCTURB OF RESCHBR'S SYSTBM
Rescher's axioms use conditional permission as the primitive operator.

All of them, except C3, have a strict implication as their main connective.
In most modal logics (those having the rule of necessitation) such a
formula is provable if and only if the corresponding formula with material
implication is provable. Therefore I state the axioms in terms of material
134 BENGT HANSSON
implication instead. C3 says something in case a formula necessarily

implies another. Like in Section VIII, I will replace strict implication by
provable material implication in order to keep the notational apparatus
within limits. The extension to the more general case presents no major
difficulties and does not affect the philosophical analysis. C3 then
becomes redundant.
The axioms are:
Cl P(t/!)
C2 P(f u g/h) +-+ P(j /h) v peg/h)
C3 if I s; g, then P(j /h) -+ peg/h)
C4 P(f () g/h) -+ P(j/g () h)
C5 P(f /h) & P(glf () h) -+ P(j () g/h)
C6 P(f It) -+ P(f /g)
C7 P(f/g) -+ P(f/k)
Alan Ross Anderson pointed out in [4] that the following formula is
provable from Rescher's axioms:
P(f /h) & P( '" fli () h) -+ P(g/h}
This is counterintuitive because "if I am permitted to smoke in the smok-
ing car, and also not to smoke in the smoking car even if I am in fact
doing so, then I am permitted to do anything whatever in the smoking
car".
Rescher answered to this by imposing a restriction on the use of C5 :
I and g have to be such that I () g is possible. The restricted C5 will be
labeled CS'.
The following comment of Anderson's should be noticed: "The
trouble appears to me to stem from CS, which enables us to arrive at the
conclusion that a contradiction might under suitable conditions be
permitted; contradictions are not 'permitted' logically, and it is doubtful
whether we should be deontically more generous than we are logically".
Rescher endorses this as "entirely correct" [14]. But the formula", P(k/f)
is equivalent to O(t/I), a dyadic counterpart to AO and BO which were
added to von Wright's system. I will therefore add this as an axiom to
Rescher's system too, in order to exclude the trivial cases where nothing
is obligatory and everything permitted. I also prefer to have the axioms

expressed in 0 ( -1-) form:
DO o (tlf)
Dl - O(klf)
D2 OU n glh) +-+ OU Ih) & o (glh)
D4 0Ulh n g) -+ OU u - glh)
DS OU u glh) & - OU /h) -+ O(g/h n - f)
DS' If f u 9 ¥= t, then OU u g/h) & - OU/h)
-+ O(g/h n - f)
D6 OU/g) -+ OU/t)
D7 OU/k) -+ OU/g)
First of all we observe that DO-D2 say that if the circumstances are
constant, we have a subsystem isomorphic to SDL. For every J we can
find a basis for the subsystem where J describes the constant circum-
stances. Let the basis be B,. The other axioms then deal with the question
how the different bases are related to each other.
THEOREM 5: The following holds in Rescher's restricted system:
(i) ifJu 9 ¥=t and B" S;; Ju 9 and not B" s;; J, then B"nN' s;; g; (ii) Bt s;; B,
for every J; (iii) B, S;; Bk for every f.
Proof.' Immediate translations of DS', D6 and D7.
THEoREM 6: In the system DO-D2, DS' the following holds for every J:
either B, is t, or B, is a one-element set or there exist disjoint bases.
Proof.' Suppose B, is not t and that x and yare two distinct elements
in B,. By Lemma 1 there is a formula 9 such that B, s;; 9 c: t and by
Lemma 2 a formula h such that x belongs to hand y to - h. BI S;; (g n h)
u (g n- h)=g¥=t. B, is not a subset of 9 nh. By Theorem SCi) B,n-(,n")
is a subset of 9 n - h. The same argument with 9 n,.., h substituted for
gnh and vice versa proves that B,n-(,n-") is a subset of gnh, which
proves the existence of disjoint bases.
COROLLARY: In the system DO-D2, DS', D6 every basis is either t or
a one-element set.
Proof: By Theorem S(ii) Bt is a subset of every basis, which excludes
disjoint bases.
136 BENGT HANSSON
THEOREM 7: The following holds in the system OO-D2, 05', 06: If

Bk=t and B, is a one-element set, then (i) B,=B, if B, £;j; (ii) B,=t if
not B, s;;; f.
Proof.' (i) Let x be the element in B, and y another element in f. By
Lemma 2 there is a 9 such that x belongs to 9 and y belongs to - g. B,
is a subset of - f u (f n g) which is distinct from t. Since B, is no subset
of - f it follows by 05' that B,n, (i.e. B,) is a subset of In g, i.e. y is
not in B,. But y was arbitrary inf - B,. Since B, is non-empty, x is the
only element in B, which then equals B t •
(ii) Suppose that B t is not a subset ofI and that B, is not t. B, is then
a one-element set and since it has B t as a subset it equals B,. B, is not a
subset of I; hence it is a subset of - f. Let x be the element in B, and y
another element in - f. There is a 9 such that x belongs to 9 and y
belongs to ,..., g. B, is a subset of lu (,..., I ng) which is distinct from t.
Since B, is not a subset of I it follows that B'n_' (i.e. B,,) is a subset of
,.., In 9 which contradicts the assumption of the theorem.
THEOREM 8: 04 is a consequence of 00-02, D5', 06-D7.
Proof.' Case 1: B, = t. By 06 every basis is t and 04 follows because
fs;;;/u-g.
Case 2: Bk is a one-element set. By 07 every basis is equal to Bk. D4
follows like in case 1.
Case 3: B, is a one-element set and Bk is t. If B"ng is t 04 is trivially
true. Otherwise B t is a subset of hng by theorem 7. Then B t is a subset
of hand B" equals B t • Then B"ng = B" and 04 is true.
We can have still more specific results if we use the unrestricted D5.
THEOREM 9: In the system 00-02, 05 the following holds for every f:
either B, is a one-element set or there exist disjoint bases.
Proof.' Suppose x and yare two distinct elements in B,. There is a 9
such that x belongs to 9 and y to '" g. B, is a subset of 9 u ,.., 9 but not
of g. Therefore B,n_g is a subset of,.., g. Permutation of 9 and", 9 yields
that B'ng is a subset of g. B,n_g and B,ng are disjoint.
COROLLARY 1: In the system 00-02, 05-D6, every basis is a one-
element set, and this set is B t •
COROLLARY 2: 07 is a consequence of 00-D2, 05-06.
COROLLARY 3: D4 is a consequence of 00-D2, 05-06.
The above theorems prove that Rescher's system is extremely strong.
In the unrestricted version all bases are equal, so circumstances never
modify obligations which is contrary to the very idea of dyadic deontic

logics. Furthermore the bases are of the type discussed at the end of
Section VIII, viz. one-element sets, which is unsatisfactory.
There are two other possibilities in the restricted version. Bt may be t
in which case all bases are t. Again the circumstances are irrelevant. Or
Bt is a one-element set and Bk is t. This is the case described in the
hypothesis of theorem 7. In every circumstance where Bt is achievable a
formula is obligatory exactly when it does not exclude Bt • In all other
circumstances nothing is obligatory except the trivial t. Although not as
horrible as the other two, this case certainly lacks the flexibility in ex-
pressing secondary duties which we want a dyadic deontic logic to have.
The corollaries to Theorems 6 and 9 result from the theorems by
adding an axiom, viz. 06, which excludes the possibility of disjoint bases.
This is all that is needed and therefore many other axioms would have
done the same job, e.g. some suggested by von Wright. The following
example, due to Castaneda's [8], is an argument against 06: If you
marry Mary Jones you ought to support her, but there is no obligation
to support her if the circumstances are only tautologous. All this suggests
that 06 should not be accepted. Castaneda also argues against 07 in [9]
on intuitive grounds.
The existence or non-existence of disjoint bases plays a certain role in
von Wright's system, which is the subject of the next section.
XII. VON WRIGHT'S AXIOMS
EI-E4 is the axiom system which von Wr!ght presented in [17]. In [18]
he changed EI to El'. The reason for this will be discussed below. EO is
the same innocuous axiom as AO, BO and DO:
EO o (t/I)
EI '" O(k/I)
El' '" O(k/t)
E2 OU n g/h) +-+ OU /h) & o (g/h)
E3 OU/g) & OU/h) -+ oU/g u h)
E4 OU /g u h) -+ OU /g) & OC! /h)
138 BENGT HANSSON
As in Rescher's system EO-E2 say that under constant circumstances

the deontic structure is as in SDL. Again we may use the notation B/
for the basis for obligations under circumstance f. The effect of the re-
placement of E1 by E1' is that we can no longer guarantee that every
basis is non-empty. E1' only says B,=Fk.
THEOREM 10: In the system EO, E1', E2, E4 the following holds for
every 9 and h: Bg u B" !;;; Bgu".
Proof: E4 says that if Bgult is a subset of j, then both Bg and Bit are
subsets of f. Now suppose that Bg contains a point x which is not in
Bgu". Bgu" is the intersection of a class of formulas. At least one of them,
say e, fails to contain x. By E4 Bg is a subset of e, which contradicts that
x is in Bg.
COROLLARY 1: BIo:!;;; B f holds for every fin the system EO, El', E2, E4.
Proof: BIo: u B f !;;; Bkuf •
COROLLARY 2: Bf !;;; B, holds for every fin the system EO, El', E2, E4.
Proof· BfuB_ f !;;; B fu - f •
If these corollaries are compared with Theorem 5, (ii) and (iii), it
becomes obvious that there is essential disagreement between the intu-
itions behind von Wright's and Rescher's systems. It is, however, inter-
esting to note that if El is accepted, then Corollary 1 implies that there
are no disjoiQ.1 bases. Thus it is impossible to add Rescher's D5' to von
Wright's system without making it forbiddingly strong and it also would
not have helped Rescher if he had made a drastic change from D6-D7
to BIo: !;;; B/ !;;; B,.
THEOREM 11: In the system EO, E1', E2-E4 the following holds for
every fand g: BfuBg=B/ug.
Proof· E3 and E4 together say that Bfug is a subset of h if and only if
both Bf and Bg are subsets of h. Therefore Bf u Bg is a basis under
circumstances fu g. The union of two theory-sets is a theory-set. But
theory-set bases are unique. Therefore Bf u Bg equals Bfug.
Now let us see why von Wright changed his mind from El to E1'. The
following formula is provable in his system: O(l/g)-+- O( - f/h). For
suppose o (I/g) and O( - f/h) are both true, i.e. that Bg is a subset off
and B" is a subset of - f. By Theorem 10 Bgu" is a subset of both Bg
and B". This is possible only if Bgu" is k, which contradicts E1.
Von Wright says about this in [18]: "It has been proved that, if there
is a 'duty to see to it that a certain window is closed should it start raining,
then there cannot be a duty to see to it that the window is open should
the sun be shining. This is manifestly absurd. Generally speaking: From
a duty to see to a certain thing under certain circumstances nothing can
follow logically concerning a duty or not-duty under entirely different,
logically unrelated, circumstances."
What von Wright wants is then the existence of (or at least the possi-
bility of the existence of) two different circumstances 9 and h and a
formula/which is obligatory under 9 and forbidden under h, i.e. BII shall
be a subset of/and Bh a subset of '" f, i.e. BII and Bh shall be disjoint!
He provides for this by changing E1 to E1'.
Does this change really exclude the kind of situations he described?
I do not think so. True, the theorem mentioned is no longer a theorem;
but there still are semantical difficulties which are not expressible in
formulas. Let us call a circumstance/ abnormal if o (kl/) is true; other-
wise it is normal. The crucial point in the proof of the unwanted theorem,
as von Wright sees it, was that 9 n h was necessarily normal. Now assume
that there is at least one abnormal circumstance and that we accept El'.
Then the set T of all complements of abnormal circumstances is a
consistent theory. For suppose that / and 9 are in T, i.e. that 0 (k/ '" /)
and 0 (k/ '" g) are true. By E3 0 (k/ '" / u '" g) is true, i.e. 0 (k/ '" (f n g»)
is true. Therefore/ng is in T. Now suppose that/is in T and that/is a
subset of g. 0 (k/ '" /) can as well be written in the form 0 (k/ '" / u( '" /
n,.., g») which by E4 implies O(k/ '" / n '" g) which is O(k/ '" g). There-
fore 9 is in T. k is not in T because t is normal by El'. Therefore T is
consistent. Let its non-empty basis be B and let x denote an arbitrary
but from now on fixed element in B. A cir~umstance J is normal exactly
when its complement is not in T, i.e. when B is not a subset of ,...., f, i.e.
when / has non-empty intersection with B. As a special case, every
formula containing x, forms a normal circumstance.
Now let / and 9 be any two formulas. One of the formulas / ng,
/ n '" g, '" / n g, '" / n,..., 9 is true in the valuation x. Therefore this
formula forms a normal circumstance. It is no longer true that every circum-
stance is normal, but there is at least one normal circumstance in· each qua-
druple of the type mentioned. If/and 9 are 'entirely different, logically
unrelated', it seems safe to say that the components of the three other for-
mulas are so too. So there is still a wealth of logically unrelated circum-
stances, such that whatever is obligatory under one of them is permitted
140 BENGT HANSSON
under the other. This is contrary to what von Wright saiti in the passage
quoted above. In general: if obligations under circumstance f are to be
logically independent of obligations under circumstance g, then f n 9 has
to be false in every valuation in B, i.e. f n 9 has to be abnormal. If von
Wright wants obligations under two circumstances to be independent as
soon as the circumstances are independent, then the intersection of two
independent circumstances must always be abnormal, which in turn
contradicts even El'.
The system EO, El', E2-E4 allows conflicting duties. This would be
a shortcoming of a logic of primary duties, but, von Wright argues, it is
a virtue in the case of a logic of conditional duties. Consider e.g. some-
one who has promised to do f, while f is forbidden. He is obligated to
refrain from f, but also to do f because of his promise. It may be that
this is a good thing taken by itself, but it is not when combined with E2,
because then everything becomes obligatory as soon as we have con-
flicting duties. And I do not think that von Wright intended to say that
someone is obligated to murder his uncle only because he promised to
punch his brother on his nose next time he caught sight of him.
But changing El to El'is not the only way to allow disjoint sets. One
can e.g. drop or change E4 instead. I think that this is a more attractive
way, since E4 seems to have some counterintuitive consequences. Let
the circumstances be that someone tries to save somebody from drown-
ing. He has succeeded in landing the man. If the man is unconscious he
is obligated to give him artificial respiration, but if the man is dead he is
not. Since only one of o (//g) and o (//h) is true, it cannot be true that
O(//g uh), i.e. if the man is dead or unconscious (and one does not know
which) there is no obligation to give him artificial respiration.
An example borrowed from Lawrence Powers' [12] also questions the
validity of E4. John Doe has got Suzy Mae pregnant. According to some
system of norms (admittedly not commonly accepted nowadays) he is
obligated to marry her under these circumstances. But, by theorem 10,
BIn, is a subset of B/ . Therefore everything obligatory under f is still
obligatory underf ng. So John Doe is still obligated to marry Suzy Mae
whatever may happen in addition to the circumstances already described.
In fact John shot poor Suzy through the head when he heard about her
condition, so he is now obligated to marry a dead girl.
Although von Wright's system admits more models than Rescher's,
it still has unpleasant consequences. I think that the system will improve
if disjoint bases are provided for in another way than changing El to El'.
How this can be done will be discussed in the next few sections.
XIII. ON THE INTERPRETATION OF 'CIRCUMSTANCES'
It is perhaps tempting to conclude from the two preceding sections that

both Rescher's and von Wright's systems are basically wrong and that a
fresh start has to be made. This, however, is not my view. Provided that
we are trying to construct a dyadic counterpart to SDL, I say that the
basic intuitions behind the two systems are sound, and, furthermore,
very similar to each other. The difficulties arise because of inessential
axioms which in some cases happen to interact strongly with other
axioms.
One question immediately presents itself. How can the two systems be
similar when there is the 'essential disagreement' mentioned in Section XII
between Theorem 5 and the corollaries to Theorem 1O? The answer is:
the theorems themselves are contrary, but the role they play in the
systems similar. They exclude disjoint bases if combined with Dl and El.
It should be clear from the discussion of von Wright's system that disjoint
bases are desirable and it should also be clear that if disjoint bases are
provided for, then the undesired strength disappears from Rescher's
system.
Who is right then? What happens to obligations when a certain
circumstance becomes true? In Rescher's system some obligations may
disappear and no new are created; in von Wright's system no obligations
disappear and some may be created. The robbery of Jones by Smith
creates the conditional obligation to restore the money. Therefore Rescher
is wrong. And if the circumstance that Smith quits his job comes true,
his employer is no longer obligated to pay him a salary. Therefore von
Wright is wrong. Obviously the situation is not simply that obligations
disappear or appear automatically when the circumstances become more
specific. D6, D7 and E4 do not seem correct. In the next section I will
present semantic foundations for dyadic deontic logics. They will preserve
all features of Rescher's and von Wright's systems, except that D6, D7
and E4 will not be valid. D 1 and El will be valid with a qualification only.
Obligations of the form 0 (llf) and 0 ( '" flf) will playa special role
142 BENGT HANSSON
in those logics. What does it mean to say O( '" Ilf)? Let/'be 'Smith robs
Jones'. It seems rather pointless to say 'Smith ought to refrain from
robbing Jones in the circumstance where he actually robs him'. If Smith
has robbed Jones, he cannot 'undo' it. He can restore what he robbed
(and this is obligatory under circumstance f in normal norm systems)
but this act is not the act of refraining from robbing Jones. We may
perhaps claim that the sentence in question only means that he should
not have done what he did, but then there would be no reason to mention
the circumstances; no matter what he actually did, he should not rob Jones.
Perhaps the situation is different iff is an act like 'Smith is smoking
in a no-smoking car'. We could then read D(......, flf) as 'if Smith is
smoking in a no-smoking car, he ought to stop'. This sounds good, but
there are some fine points here. Let us discuss this and other proposed
readings of 0 ( . . . , flf) systematically.
First, dyadic obligations are secondary, reparational obligations, telling
someone what he should do if he has violated (intentionally or not) a
primary obligation. Therefore they should not merely say that the agent
should not have done what he did; the primary obligation 0 ( ......, f)
already said that and the situation would be completely described by the
mixed formulaf &O( '" f) if one wants to stress that the agent actually
violated the 9bligation. If one takes conditional obligations seriously,
one has to realize that an agent cannot 'undo' what he has actually done.
The best way to read O(flg) is then 'now that the agent actually has
done g, he ought to do f'.
Second, if we return to the smoking Smith, it seems to make sense to
say 'now that Smith actually has smoked in a no-smoking car, he ought
to refrain from smoking in a no-smoking car', because it does not
necessarily mean that he shall 'undo' the smoking already done, but only
that he shall refrain from further smoking. More carefully stated the
obligation then reads: 'now that Smith has smoked in a no-smoking car
up to this moment he ought to refrain from continuing after this moment'.
But this is not an obligation of the form 0(......, Ilf).
From this I conclude that formulas like O(I/g) shall never be true if
f and g are disjoint, if circumstances are taken seriously. And by this I
mean that the circumstances are regarded as something which has actually
happened (or will unavoidably happen) and which cannot be changed
afterwards.
XIV. THE DYADIC SYSTEMS DSDL1, DSDL2, AND DSDL3
The obligations and permissions of SDL may be explained like this:

Certain possible worlds are ideal. You shall always try to make the real
world an ideal world. Some formulas are true in every ideal world.
Therefore you have to make these formulas true if the world is to become
ideal. These formulas are called obligatory. Formulas which are true in
at least one ideal world (though not necessarily in all of them) are called
permitted and you may make them come true if you want that ideal
world to come true. Permitted formulas are in general not obligatory
because you may just as well pick another ideal world as the one you
want to realize. You shall not make true formulas which are false in
every ideal world because then no ideal world can be realized. Such
formulas are called forbidden.
The problem of conditional obligation is what happens if somebody
nevertheless performs a forbidden act. Ideal worlds are excluded. But it
may be the case that among the still achievable worlds some are better
than others. There should then be an obligation to make the best out
of the sad circumstances. The following seems to be a rather straight-
forward way to generalize the semantics of SDL to the dyadic case:
If some circumstance is given, some a priori possible worlds are ruled
out while others are still achievable. Among the latter ones there may be
one or several which are at least as ideal as all other achievable possible
worlds. You shall always try to make one of them come true. Some
formulas are true in all of these most ideal worlds. Therefore you have
to make these formulas true if the world is to become maximally ideal.
Such formulas are called obligatory under the given circumstance. Etc.
It is now easy to construct a concept of validity along these lines. A
relation R will reflect the relation 'is at least as ideal as'. The relation P
(,strictly more ideal than') is defined by: xPy=both xRy and not yRx.
Now, x is R-maximal in the setJifthere is no y inJsuch that yPx holds.
If Rand R' are two relations and P and P' derived from them and if
P=P' then an element is R-maximal in a certain set if and only if it is
R'-maximal in the same set. In particular, if R' is the reflexive closure
of R, then the two concepts of maximality coincide. Therefore I will
restrict myself to reflexive relations in the following.
A valuation for the first dyadic standard deontic logic (DSDLl) is a
144 BENGT HANSSON
reflexive relation R defined on the set t of all valuations of the BL for

DSDLI. A DSDLl-formula of the type OU/g) is true in the valuation
R if and only iff contains all R-maximal elements in g. Other formulas
take truth values according to the rules or propositional logic. A DSDLI
formula is valid if and only if it is true in all DSDLl valuations.
DSDLI is a rather weak system. Since it may be the case that some
formulas contain no maximal elements, there may exist empty bases, i.e.
circumstances under which everything is obligatory and nothing per-
mitted. DSDL2 is a stronger system where this does not happen except
when the circumstance is k.
A DSDL2 valuation is a reflexive relation R defined on t and having
the property that every non-empty formula contains at least one R-
maximal element. Truth and validity are defined as in the case of DSDLI.
DSDL2 is also rather weak. Stronger systems arise if one places more
conditions on R. A very natural condition seems to be that R fulfils the
usual axioms of preference relations, i.e., that it is transitive and total,
because R is meant to reflect some kind of preference.
A DSDL3 valuation is a DSDL2 valuation which is also transitive
and total (complete, strongly connected). Truth and validity are defined
as in the case of DSDLI. DSDL3 is the system which is most similar to
ResCher's and von Wright's.
xv. VALID AND INVALID FORMULAS
(1) o (t//)( = DO, EO).

Valid in all three systems.
(2) 0U/f).
Valid in all three systems. This is in a way a counterpart to (1), because
one may think of obligations relative to circumstance f as obligations in
a restricted universe: of all the a priori possible worlds only those in f
are now achievable; therefore f plays the role of the universe and (2)
only says that at least something is obligatory under f.
(3) '" o(k//)( = Dl, El).
Not valid in any of the systems. The only counterexamples in DSDL2
and DSDL3 are those where f is k. If (3) is considered desirable we may
AN ANALYSIS OF SOME DEONTIC LOGICS 14S
of course redefine the truth of 0 (flk). Our intuition on this matter is so

weak that no redefinition would be counterintuitive. But such a redefini-
tion may make it necessary to put in qualifying clauses in other valid
formulas or complicate the formal apparatus in other ways. Since the
question has almost no philosophical significance, I choose the solution
which makes the system most uniform.
(4) '" o (kjt)( = El').

Valid in DSDL2 and DSDL3 but not in DSDLl.
(5) O(j (\ gjh) +-+ O(j jh) & o (gjh)( = D2, E2).
Valid in all three systems. This formula reflects the idea of a basis and
since the DSDL systems are constructed as dyadic extensions of SDL it
should be valid.
(6) O(jjh (\ g) -+ O(j U '" gjh){= D4).

Valid in DSDL2 and DSDL3 but not in DSDLl.
(7) O(j u gjh) & '" o (jjh) -+ O(gjh (\ '" f)(= DS).
Valid in DSDL3 but not in DSDLl or DSDL2. This is the unrestricted
version of Rescher's fifth axiom. Anderson's problem in this connexion
is solved by the considerations in Section XIII.
(8) O(j u gjh) +-+ O(j jh) v O(gjh (\ '" f).

Valid in DSDL3 but not in DSDLl or DSDL2. The implication to the
right is of course formula (7) and the implication to the left a combination
of (6) and (S). This formula is interesting because it is equivalent to
P(f (\ glh)+-+P(flh) &P(gl/ (\ h) which is the essential axiom in von
Wright's first dyadic deontic logic in [16].
(9) O(j jg) & O(j jh) -+ O(jjg u h)( = E3).

Valid in all three systems.
(10) O(j jg u h) -+ O(j jg) & O(j jh)( = E4).
Invalid in all three systems.
(11) O(jjg) -+ O(jjg (\ h).
146 BENGT HANSSON
Invalid in all three systems. (11) is a consequence of (10) and should be

compared with (12).
(12) o (f/g) & '" 0(- h/g) --. O(f/g ("\ h),
or equivalently
o (f/g) &P(h/g) --. O(f/g n h).
Valid in DSDL3 but not in DSDLl and DSDL2. An obligation remains
an obligation if one does something permitted.
(13) O(f/g u h) --. O(f/g) v O(f/h).
Valid in DSDL3 but not in DSDLl and DSDL2. This is a weaker version
of (10).
(14) O(f/g) --. o (flt)(= D6).
(15) O(f It) --. O(f Ig)( = Theorem 10).
(16) O(f/k) --. o (f/g)(= D7).
(11) o (f/g) --. O(f/k).
Valid in all three systems, but only because everything is obligatory under
circumstance k.
(18) O(glf) & o (h/J n g) --. O(g n hI!).
Valid in all three systems. This formula says that if it is obligatory to do
something and also obligatory to do something else when the first thing
is done, then it was obligatory to do both things in the first place. Rescher's
fifth axiom when expressed in P, C5, is the counterpart of (18) with P
substituted for O.
(19) O(f n g/h) --. O(f/g n h).
Valid in DSDL3 but not in DSDLl and DSDL2. If you should do two
things and you do one of them, you still have the other one left. This is
a counterpart to C4.
University of Lund
BIBLIOGRAPHY
[1] A. R. Anderson, 'The Formal Analysis of Normative Systems', in The Logic 0/

Decision and Action (cd. by N. Rescher), The University of Pittsburgh Press,
Pittsburgh, 1967, pp. 151-213.
[2] A. R. Anderson, 'A Reduction of Deontic Logic to Alethic Modal Logic', Mind
67 (1958) 1~103.
[3] A. R. Anderson, 'The Logic of Norms', Logique et analyse 1 (1958) 84--91.
[4] A. R. Anderson, 'On the Logic of Commitment', Philosophical Studies 10 (1959)
23-27.
[5] A. R. Anderson, 'Reply to Mr. Rescher', Philosophical Studies 13 (1962) 6-8.
[6] A. R. Anderson, 'Some Nasty Problems in the Formal Logic of Ethics', Nous 1
(1967) 345-360.
[7] L. Aqvist, 'Good Samaritans, Contrary-to-Duty Imperatives and Epistemic
Obligations', Nous 1 (1967) 361-379.
[8] H.-N. Castafteda, 'The Loaic of Obligation', Philosophical Studies 10 (1959) 17-23.
[9] H.-N. Castafi.eda, 'Correction to "The Logic of Obligation" (A Reply)" Philo-
sophical Studies 15 (1964) 25-28.
[10] R. M. Chisholm, 'Contrary-to-Duty Imperatives and Deontic Logic', Analysis 13
(1963) 33-36.
[11] S. Kanger, New Foundations For Ethical Theory, Stockholm 1957. Reprinted in
this volume, pp. 36-58.
[12] L. Powers, 'Some Deontic Logicians', Nous 1 (1967) 381--400.
[13] N. Rescher, 'An Axiom System for Deontic Logic', Philosophical Studies 9 (1958)
24--30.
[14] N. Rescher, 'Conditional Permission in Deontic Logic', Philosophical Studies 13
(1962) 1-6.
[15] G. H. von Wright, 'Deontic Logic', Mind 60 (1951) 1-15.
[16] G. H. von Wright, 'A Note on Deontic Logic and Derived Obligation', Mind 65
(1956) 507-509.
[17] G. H. von Wright, 'A New System of Deontic Logic', Danish Yearbook 0/
Philosophy 1 (1964) 173-182. Reprinted in this volume, pp. 105-115.
[18] G. H. von Wright, 'A Correction to a New System of Deontic Logic', Danish
Yearbook 0/ Philosophy 1 (1965) 103-107. Reprinted (in part) in this volume,
pp. 115-120.
[19] G. H. von Wright, 'Deontic Logics', American Philosophical Quarterly 4 (1967)
136-143.
NOTE
• Reprinted by permission of the author and the publisher from Nous 3 (1969) 373-398.
KRISTER SEGERBERG
SOME LOGICS OF COMMITMENT AND OBLIGATION
The topic of this paper is the logical relationship between the notions of
commitment and obligation. When this question was first raised it was
thought that the former notion could be analysed in terms of the latter
and the Boolean connectives of classical logic. However, in the end
almost every effort to carry out such an analysis has led to intuitively
unacceptable consequences, and there would seem to be substantial
agreement among today's logicians that the notion of commitment
cannot be reduced to that of obligation. 1 In this paper we define a family
of propositional logics in which both commitment and obligation are
formalized as independent operators.
The words 'commit' and 'commitment' have several uses in English
only some of which are of interest to deontic logic. Even among the latter
there is an obvious distinction to make. In one sense commitment is a
one-place concept; in this sense, we believe, 'commitment' is to be
identified with 'obligation'. In another sense - its 'proper' sense - it is a
two-place concept which is rendered by '(the fact) that A commits the
agent to (the proposition that) B'. (Actually, it may be argued that both
concepts of commitment ought to be relativized to agents in order that
their full logical form be brought out. According to this view commitment
in the first sense is an at least two-place concept, and 'proper' commitment
is at least four-place: 'By virtue of (the fact that) A, g commits b to the
proposition that B'. In this paper, though, neither g nor b will play a
significant role. More precisely, g will be entirely neglected and b will be
suppressed; unless we specify otherwise it may be assumed that b is an
arbitrary fixed agent.)
We shall use A com B for the proposition that A commits the agent
to B. Another way of expressing the same proposition would be to say
that if A then the agent is committed to B, but then it must be noted that
'committed' is used in the sense of 'obligated' and that the 'if-then'
cannot be identified with material implication.
R. Hilpi""" (ed.). Deo"tic Logic: Introductory and Sy.tellJfJtic Relllllng•• 148-158. All right. reserved.
CopyrlghtO 1970 by D. Reidel Publishl"g Comptmy. Dordrecht-Holland.
SOME LOGICS OF COMMITMENT AND OBLIGATION 149
The notion of similarity between moral situations is of fundamental

importance to ethics; in fact, without it ethics, as we know it, would be
impossible. We shall make use of this notion in our analysis of commit-
ment. Suppose somebody maintains that a certain agent is committed
to B if A obtains - in our symbolism, A com B. For someone who wants
to challenge this contention there is something like a method available:
indicate a situation similar to that referred to by the first speaker such
that A is true in that situation and yet B is not obligatory for the agent
(or his counterpart in the new situation). To be sure, this method is not
an algorithm, but if a situation of the kind envisaged is found then the
claim that A COIB B has been refuted. Note that it is not usually enough
to consider only the situation first referred to, for the contention A co. B
may be challenged even though A is false in that situation or the aaent is
committed to B. On the other hand it would usually be a mistake to
consider every possible situation. To illustrate this point by a somewhat
silly example, let A be "Mr. Ahl utters 'I do' in earnest" and let B be
"Mr. Ahl marries Miss Bj6rk". Suppose that u is a possible situation in
which Miss Bj6rk has just said in earnest to Ahl, eligible bachelor, "if
you ask me 1 will marry you". No doubt A COIB B is true in u; under the
circumstances, saying "I do" in earnest will commit Ahl to marrying the
girl. Now there are infinitely many possible situations in which Ahl is
appearing in court as a witness and is being asked whether he solemnly
swears to tell the truth etc., to which he responds by uttering, in earnest,
"I do". Tn one of those situations, call it w, Ahl has not only not promised
Miss Bj6rk to marry her, he has never met her, and in fact he is already
married to the former Miss Ceder. In w, then, Ahl is under no obligation
to marry Miss B,i6rk. Hence, whereas A is true in w, B is not obligatory.
But obviously the existence (in some sense) of the possible situation W
is irrelevant when we want to consider the truth of A COIB B in u: u and
ware too dissimilar. Figuratively we might say that, at least from the
point of view of evaluating A com B in u, W is too far removed
from u.
To characterize the similarity relation would be a difficult and perhaps
elusive task,a but it does not seem arbitrary to postulate its existence. To
take one more example, legislators certainly believe in, or are committed
to the belief in, the existence of similarity; for an important part of law-
making consists in describing situations, which is nothing else but
150 KRISTER SEGERBERG
singling out classes of individual possible situations between which some

sort of similarity obtains, or which are all similar to some archetypal
situation. (From this remark it may be thought that similarity (with
respect to certain features) should be regarded as an equivalence relation,
but that may well be disputed, and we shall not here take it for granted
that it is.)
A possible situation may be similar to another with respect to certain
features, and not similar with respect to others. Of two possible situations
that are both similar to a third, one may be more similar to the third than
the other is. Suppose it were meaningful to talk about the 'deontic
distance' between possible situations. Then two situations would be
similar to some degree whenever the deontic distance between them
would be finite; the deontic distance between two situations would be
zero if and only if they would be completely similar from a deontic point
of view (that is, with respect to every set of deontically relevant features);
a situation u would be more similar to a situation w than a situation v
would be if and only if the deontic distance between u and w would be
smaller than that between v and w; and so on.
The idea of 'deontic distance' is of course a rather fanciful one, and
we shall not attempt to introduce anything like a deontic metric on the
space of possible situations. However, we shall suggest a way of capturing
at least partially the qualitative aspect of such an idea. For this purpose
we introduce the key notion of this paper, that of 'neighborhood'. A
neighborhood is simply a set of possible situations. In general not all
sets of situations are neighborhoods, but we shall assume, for each
possible situation u, that there is a set Nu of (deontic) neighborhoods
around u. Our intuitive background is this: if ex is a neighborhood around
u then each possible situation in 0( has at least some similarity to u; a
situation that belongs to every neighborhood of u will have to be com-
pletely similar to u in all deontically relevant aspects; if 0(, Pare neighbor-
hoods around u and 0 C 0( C p, then some situations in 0( are more
similar to u than those in p - 0(; and so on (cf. the concluding sentence of
the preceding paragraph). The problem of finding the set of neighbor-
hoods around a concretely given situation is not dealt with in this paper.
Here we shall take the existence of neighborhoods for granted and see
how this idea leads us to a logic of commitment. For now we are able to
express our intuitions regarding 'proper' commitment with greater
precision than before: A com D shall be true at u if and only if throughout

some neighborhood around u, whenever A is true then B is obligatory.
Evidently this theory of commitment depends on our theory of obligation
and on what we take our neighborhoods to be. For this reason we get
a family of logics of commitment rather than just one logic of commit-
ment. This point will become apparent in the formal development of the
theory, to which we now turn.
It is not clear to the author who deserves the credit for originating the
'neighborhood' semantics, but the formal idea has been available for
some time (see Montague [7] and Scott [9]). For his own part the author
became interested in its possible application to philosophical problems
by reading David Lewis's one page draft [6].
II
We consider a language for propositional logic having ..1 (falsity) and

-+ (material implication) as primitive Boolean connectives and 0 (obli-
gation) and com (commitment) as primitive non-classical operators. We
use the familiar abbreviations -', A, V, and +-+; T will be short for the
<
formula ..1-+..1. By a model we understand a structure U, R, N, V)
where U is a set called the domain of the model (intuitively its elements,
sometimes called points or possible worlds, are possible situations); R is
a binary relation on U (the alternativeness relation that goes with 0);
N is a function on U such that, for every UE U, Nu is a set of subsets of U
(the family of neighborhoods ofu); and Vis a function assigning to each
propositional letter P a subset V(P) of U (the valuation of the model).
We shall use u, v, W, ••• for elements of domains, IX, p, 'Y, ••• for neighbor-
hoods, and A, D, C, ... for formulas.
Truth of a formula at a point u in a model t:ft = <U, R, N, V), denoted
by 1= ~, is defined as follows.
1 For propositional letters P, 1= ~ P iff (if and only if) u E V (P) .
3 1=: A -+ D iff I=~ A only if I=~B.

4 F: OA iff for every v such that uRv, F~ A.

5° F: A com B iff there exists some IX EN" such that for all v E IX, if
F~ A,then F! B holds for all w such that vRw.
Throughout the paper we shall use the abbreviation IICII'- for the exten-
sion ofC in tW, that is, for the set {U:UE U &F~C}. Note thatthe clause SO
is equivalent to the condition that A com B if and only if there is some
neighborhood IX of u such that IX £; IIA -+ OBII'-.
We say that A is true in a model if F"A for every u in the domain of the
model. The set of formulas true in all models can be axiomatized as
follows. Let our rules of inference be the following three:
A A-+B
R1 (Modus Ponens)
B
A
R2 (Obligation)
OA
(A -+ OB) -+ (A' -+ OB')
R3°
A com B -+ A' com B'
Our axioms are the tautologies of classical propositional logic plus all
instances of the familiar schema
*0 O(A -+ B) -+ (OA -+ OB).
Let this basic calculus be called BO. It can be shown, for example with
the now so common Henkin type of argument, originally used by
Makinson and Scott, that a formula is demonstrable in BO if and only
if it is true in every model.
From the point of view of our intended interpretation of 0 as obli-
gation and com as commitment this calculus seems too weak. It is stan-
dard among deontic logicians to demand that falsehood not be obligatory.
Moreover, it seems to this author that one cannot help being committed
to what is logically true, although no doubt some philosophers would
disagree. Finally, one of the basic facts about the kind of commitment
we have in mind is that if A commits one to B and if A is the case, then
one is committed - in the sense of being obligated - to B (cf. Anderson [1]

and Aqvist [2]). But of the following schemata none has all its instances
demonstrable in the calculus BO:
*1 0.1 -+ .1
*2 AcomT
*3 A com B -+ (A -+ OB).
Let F O be the calculus obtained by adding all instances of :#= 1-3 to *
the stock of axioms of BO, keeping the three rules as before. We regard
FO as a fundamental deontic logic of obligation and commitment. It is
not difficult to show that a formula is demonstrable in pO if and only if
it is true in all models that satisfy these conditions:
(i) 'V u 3 v (uRv).

(ii) 'Vu(Nu #= 0).
(iii) 'V u (if Nu #= 0, then u E n Nu).
The first condition is well known from traditional deontic logic. The
second says that every element has at least one neighborhood, the third
that every point belongs to each of its neighborhoods. How reasonable
are these conditions? Probably (ii) is the most controversial one. To
deny (ii) would amount to accepting that there are situations which are
not similar to any possible situation, not even themselves. This should
be compared to denying (i) which amounts to accepting that there are
situations which have no alternatives. Evitiently both (i) and (ii) serve
to rule out the existence of very singular situations. While models satis-
fying (i) and (ii) are singled out for study here, this is not to say that
models violating (i) or (ii) are void of interest.
There are many other model theoretic conditions one should like to
discuss. It soon becomes evident, though, that for the purpose of such
discussions our choice of primitives is inconvenient; the operator com
is too complicated. Already the formulation of clause 5° in the definition
of truth above reveals that we could introduce in place of com a new
unary primitive [J and replace 5° by the clause
5 1= ~ [J A iff there exists some 0( E Nu such that 0( £; IIA II"

The point is that com may be defined in terms of c:J and the other primi-
tives:
A comB = df c:J (A ~ OB)
(Note how well this fits with reading 'A com B' as 'if A then the agent is
committed to B'!) With the new set of primitives we obtain a more
elegant axiomatization of the set of formulas true in all models. Let B
be the basic calculus whose axioms are the tautologies of classical logic
*
and the instances of schema 0, and whose rules are Modus Ponens,
Obligation, and the new rule
A~B
R3
One can prove that the set of formulas true in every model is exactly the
set of formulas demonstrable in B.
The new language is richer than the old one, for whereas com can be
defined in the new language, 0 cannot be defined in the old. However,
unlike com, the operator 0 does not seem to have an obvious counter-
part in a natural language like English. In view of our intuitive remarks
in the introduction one would like to read [JA as 'it holds deontically
essentially that A' or something of this sort. But for such a reading of 0
the calculus B is not strong envugh; at least we must add as axioms to B
*
all instances of 1 and of
In this way we obtain a calculus F which corresponds to FO as B corre-

sponds to BO. One can prove that a formula is a theorem of F if and
only if it is true in all models that satisfy the conditions (i)-(iii) above.
In the light of our intuitions, even conditions (i)-(iii) seem too weak.
If the approach presented in this paper is fruitful, it will be important to
ask what other conditions should be imposed on models in order to
characterize the concept of commitment. Or, with a more careful formul-
ation, one must study the effect of imposing plausible conditions on
models and see whether any of the corresponding concepts of commit-
ment approximate those found in nature. We shall mention only a few
such conditions here.
Consider the condition

(iv) 'V u(if ex, fJ e Nil' then ex n fJ eN,,)
and the schema
=#: 4 8 A A 8 B --+ 8 (A A B).
Let FK be the calculus got be adding all instances of =#:4 as axioms to F.
Then the set of formulas demonstrable is identical with the set of formulas
true in every model satisfying conditions (i)-(iv). It makes no difference
to the resulting logic if instead of (iv) we add to (i)-(iii) the condition
(iv') 'V u (N" is a filter) .
(N" is a filter if and only if for all ex, fJ, ex nfJeN" if and only if exeN" and
fJeN".) Another condition equivalent in the present context to (iv) and
(iv') is
(iv") 'V u (if N" =1= 0, then n N" eN,,).
Schema =#: 4 and the conditions that correspond to it seem reasonable.
One may note the following technically interesting fact. In the presence
of both =#: 2 and =#: 4 the operator 8 behaves like a Kripke type necessity
operator. In fact we could define a new concept of model <U, R, S, V),
called Kripke model, where U, R, V are as before and S is another
alternativeness relation, that is, a binary relation on U .Thus, here we have
two alternativeness relations, one that tells us whether one possible
situation is a deontic alternative of another, and one which tells us
whether, from the deontic point of view, one possible situation is com-
pletely similar to another. In view of (ii) and (iv") the connection between
the two kinds of models is direct: if for all u, ve U
uSv iff v en N",
then clearly <U, R, N, V) and <U, R, S, V) are equivalent in the sense
that exactly the same formulas are true in them.
Two interesting schemata mixing 8 and 0 are
=#: 5 OA --+ 80A
=#: 6 ..., OA --+ 8 ..., OA.
One can show that a formula is demonstrable in the calculus obtained
* *
by adjoining all instances of 5 and 6 to FK if and ('nly if it is true
in every Kripke model satisfying condition (i), the condition that S is
reflexive and the additional conditions
(v) V u V v V w(ifuSv and vRw then uRw).
(vi) V u V v V w(if uSv and uRw then vRw).
The schemata
*.7 'O'OA-A.
*8 OA-OOA
are well known; their 'corresponding' conditions a,re
(vii) S is symmetric.
(viii) S is transitive.
The set of formulas demonstrable in the calculus obtained by adding all
instances of ** 5-8 to FK is of course identical with the set of formulas
true in every Kripke model in which (i), (v), (vi) are satisfied and S is an
equivalence relation. Remembering that the set of S-alternatives of a
point may be thought of as a neighborhood of that point - in fact the
smallest one -.we observe that those models have the following interesting
property: for every point u there is a neighborhood such that throughout
the neighborhood exactly the same obligations and the same commit-
ments hold as in u. Thus such a model is partitioned into equivalence
classes within each of which non-deontic facts may vary but not deontic
ones. If one accepts the underlying logic one may perhaps say that this
concept of model explicates the concept of a deontic situation. 3
III
Although there may be many objections to the theory of commitment

presented in this paper we shall deal with only one here. This objection,
due to Lennart Aqvist, concerns the fact that already in BO we have as a
theorem schema
AcomC -(A A B)comC.
Is this consonant with our intuitions? If Nilsson's car breaks down and
Ek promises to lend him his for the day we may say that Ek is committed
to giving Nilsson the car when he comes to pick it up, but few would say
that such a commitment obtains if it turns out that Nilsson shows
up drunk. It must be admitted that none of our logics can handle such
cases.
The most natural reply is, perhaps, that at least one of the com-operators
in Aqvist's example is of a kind which our semantics does not formalize.
Let the exacting notion of commitment we have analysed be called 'strong
commitment'. Aqvist's example, it seems, involves a kind of 'commit-
ment-reasonable-in-principle', or 'weak commitment' for short. That
A 1\ B strongly commits me to C if A does, seems acceptable - as accept-
able as that there are cases when A, but not A 1\ B, weakly commits me
to C. If so, Aqvist's interesting observation ceases to be an objection
against our theory.
It would be tempting to identify strong and weak commitment with
absolute commitment and prima facie commitment, respectively.4 The
development of semantics for weak or prima facie notions of commitment
within the general framework defined above seems like an interesting
problem, but in this paper it is left open.
Abo Academy
BIBLIOGRAPHY
[1] Alan Ross Anderson, 'On the Logic of "Commitment"', Philosophical Studies 10
(1959) 23-27.
[2] Lennart Aqvist, 'A Note on Commitment', Philosophical Studies 14 (1963) 22-25.
[3] Brian F. Chellas, The Logical Form of Imperatives, Perry Lane Press, Stanford,
Calif., 1969.
[4] Risto Hilpinen, 'An Analysis of Relativised Modalities', in Philosophical Logic (ed.
by J. W. Davis et al.), D. Reidel, Dordrecht, 1969, pp. 181-193.
[5] Jaakko Hintikka, Models for Modalities. Selected Essays, D. Reidel, Dordrecht,
1969.
[6] David Lewis, 'Semantic Analysis of Subjunctive Conditionals', Xerox copied,
V.C.L.A., 1968.
[7] Richard Montague, 'Pragmatics', in Contemporary Philosophy. A Survey. I. Logic
and Foundations of Mathematics (ed. by R. Klibansky), La Nuova Italia Editrice,
Firenze, 1968, pp. 102-122.
[8] Dana Scott, 'A Logic of Commands', mimeographed, Stanford University,
Stanford, 1967.
[9] Dana Scott, 'Advice on Modal Logic', in Philosophical Problems in Logic: Some
Recent Developments (ed. by Karel Lambert), D. Reidel, Dordrecht, 1970, pp. 143-
173.
NOTES
lOne important exception to this statement is Jaakko Hintikka's article in [5],

'Deontic Logic and Its Philosophical Morals'. In that paper an account, in terms of
obligation and Boolean operators only, is given of two notions of commitment,
absolute commitment and prima facie commitment. Hintikka avoids the intuitively
unacceptable consequences previous analyses in such terms have led to by distinguishing
between these two notions. (Cf. also the paper by Hintikka in this volume, pp. 59-104.)
2 For an interesting solution to this problem in a somewhat related context, see Hil-
pinen [4].
3 A continued discussion of our approach to commitment would probably become
more meaningful in the context of a more comprehensive study of deontic logic such
as that attempted in Scott [8] and Chellas [3].
4 a. the paper by Hintikka referred to in Note 1.
GEORG HENRIK VON WRIGHT
DEONTIC LOGIC AND THE THEORY

OF CONDITIONS·
Deontic logic was, in origin, an off-shoot of modal logic. It got its decisive
impetus from observations of some obvious analogies between the modal
notions of necessity, possibility, and impossibility on the one hand and
the deontic or normative ideas of obligation ('ought to'), permission
('may'), and prohibition ('must itot') on the other hand. In a broad sense,
both groups of concepts can be called modal; the members of the first
group are sometimes referred to as alethic, those of the second group as
deontic modalities. (A preferable alternative to the term 'alethic' is
perhaps the term 'anankastic'.)
Beside analogies and similarities, however, there are also a number of
striking dissimilarities between the two types of modalities. Many of the
problems which have beset deontic logic since its birth are related to
these discrepancies. One difference is the absence in deontic logic of an
analogue to the principle 'Np -+ p' of modallogic.1 That which necessarily
is the case is also as a matter of fact the case; but that which ought to be
the case is far from being always actually the case. Another formal
difference between modal and deontic logic is that, whereas it is obvious
that the tautology necessarily is true ('Nt'), it is not intuitively clear that
the tautology also ought to be true ('Ot'). The idea expressed by COt' does
not seem to make good sense. By contrast, ',.., 0 ,.., t' seems not only
to make sense, but also to be true. This formula says that a contra-
dictory state of affairs is not a state which ought to be the case. This
is an analogue to the principle ',.., N,.., t' of modal logic which can
also be written 'Mt'. The principle '", N,.., t' is a weaker form of the
principle 'Np -+ p' in as much as the first follows logically from the
second (and principles of ordinary propositional logic, PL), but not
vice versa.
One can display these analogies, and failures of analogy, in the follow-
ing table:
R. Hilpinen (ed.), Deont/c Logic: Introductory and SYltematlc Reodlng8, 159-177. A.1l right8 re8er.ed.
Copyright C 1970 by D. Reltkl Publl.lhlng Company, Dort/recht·Holland.
Modal logic (System M) Deontic logic
Al N(p&q)~Np&Nq O(p & q) ~ Op & Oq
A2
A3 Nt Not: Ot
There is a further noteworthy difference between the two logics which

attracts attention: In modal logic, the interdefinability of the ideas of
necessity and possibility through the schema 'M' = df' ' " N ~' provokes
no serious objection. But the corresponding schema or equivalence in
deontic logic is by no means unproblematic. It seems feasible to admit a
'weak' notion of permittedness,2 according to which something may be,
if and only if it is not the case that the contradictory of this thing ought
to be. The dual of the formula 'O(p&q)~Op&Oq', i.e. the distribution
principle 'pep v q)~Pp v Pq', holds good of this notion of permitted-
ness. This, however, does not correspond to the way in which permission
is normally thought to be distributive over alternatives. If we are told
that we may do this thing or that thing, we normally understand this to
mean that we may do the one thing but also the other thing. The distri-
bution principle, in other words, would seem to be 'pcp v q)~Pp&Pq'.
But this principle goes with a different idea of permittedness from the
one which obeys the interdefinition schema 'P'=dr'''' 0 ~'. We can call
it a notion of strong permission. It is related to possibility (freedom) of
choice between alternatives.
II
In my paper I shall propose a somewhat novel conception of deontic

logic. In the light of this conception the discrepancies between deontic
and modal which we mentioned in the preceding section are placed in a
deeper perspective. A prospect is opened for a solution to many difficulties
of a logical and philosophical nature associated with the very idea of a
'logic of norms'.
This new conception regards deontic logic, not immediately as an
analogue of modal logic, but as a fragment of a more comprehensive
logical theory which I shall call the Logic of (Sufficient and Necessary)
DEONTIC LOGIC AND THE THEORY OF CONDITIONS 161
Conditions. Since, however, this Logic of Conditions - as I conceive of

it here - is itself a fragment of modal logic, it will a fortiori be true to say
the same of deontic logic.
To put it quite briefly, the main contention of this paper is as follows:
To say that something ought to be, or ought to be done, is to state that
the being or doing of this thing is a necessary condition (requirement) of
something else. To say that something may be, or may be done, again
has two different, typical meanings. Either it is simply a denial of the
statement that the contradictory of this thing ought to be (done), i.e. is
a necessary requirement of something else. Or it is an affirmation to the
effect that the being or doing of the thing in question is a sufficient
condition (guarantee) of something else. When a statement of permitted-
ness has the first meaning, it can also be rendered as a 'need not'-state-
ment. When it has the second meaning it is frequently couched as a
'can'-statement. I shall refer to the two meanings as the weak and the
strong (meaning of) 'may' respectively.
In calling the proposed conception of deontic logic 'somewhat novel',
I have in mind the existence of an earlier attempt in the same direction.
This is A. R. Anderson's well-known reduction of deontic to alethic
modal logic. 3 Substantially, the present paper is a further development
and extension into new dimensions of ideas originally put forward by
Anderson.
III
A satisfactory logical theory of conditions is still very much of a desider-

atum. We cannot attempt to satisfy it here. But we must agree about
some of the features of this theory before we can proceed to considering
its connexion with deontic logic.
A theory of conditions can conveniently be built in stages. On the
lowest stage, the terms of the conditionship-relation are propositions
(or some 'proposition-like' entities). These are not further analyzed, but
treated as 'wholes'. The laws of 'classical', two-valued propositional logic
(PL) are accepted for them. What else is needed? At least, I think, the
modal ideas of necessity, possibility, etc. and some minimum assump-
tions about their logic. A plausible minimum seems to be to accept the
system of modal logic known as System M (or T). Its axioms were listed
above, in Section I. The system is the common core of most modal logics
which are known and studied; it embodies the unproblematic, more or

less universally 'received' principles of modality.4
We tentatively suggest the following definition of the notion of a
necessary condition, strictly speaking of the phrase 'the truth of the
proposition that p is a necessary condition of the truth of the proposition
that q':
'Nc(p, q)' = df 'N(q -+ p)'.
Thereupon we suggest the following definition of the phrase 'it ought

to be the case that p':
'Op' = df 'Nc(p, I)'.

That something ought to be the case, or is 'obligatory' in a loose and
wide sense of that term, thus means according to our suggestion that the
thing in question is a necessary condition of a certain thing (proposition,
state of affairs) 'J'. This is a propositional constant, not a variable. What
this 'I' is, its content, we leave for the time being open.
On the above basis it is easy to prove 'O(p&q)-Op&Oq', or the
'received' distribution principle for the deontic O-operator.
'Op -+ p' is not a theorem of our theory (of conditions). This is as we
want to have it: It must not follow logically from the fact that something
ought to be that this thing is.
'", 0 '" t' is not a theorem either. This conflicts with the traditional
conception of deontic logic. A way out of the conflict is opened by the
observation that the only thing of which a contradictory state of affairs
can be a necessary condition is - another impossible state. '", 0 '" t'
would be provable, if we modified our definition above of the notion of
necessary condition by stipulating that that of which something is a
necessary condition must not itself be impossible.
With or without this additional stipulation, however, 'Ot' is a theorem.
This too is at variance with 'traditional' conceptions. As noted in
Section I, 'Ot' is not particularly wanted as a theorem of deontic logic;
~ may even wish to reject it. We can get rid of the unwanted result by
stipulating that that which is a necessary condition of something must
not itself be necessary.
The two stipulations - the one which secures theoremhood for' '" 0 '" t'
and the one which excludes 'Ot' from the status of a theorem - are em-
bodied in the following modified definition:
'Nc(p, q)' = de 'N(q ~ p) & M,.., p & Mq'.
Let the symbol 'C' signify contingency. If we accept the modified
definition of a necessary condition, we can prove the theorem 'Nc(p, q)
~ Cp&Cq'. From our definition of 'ought' we then also prove 'Op~ Cp'
and, since contingency entails possibility, 'Op~Mp'. The last may be
regarded as a version of the principle, commonly associated with the
name of Kant, that 'Ought implies (entails) Can'.
I shall call the addition 'M,.., p&Mq' to the definition of Nc a contin-
gency-clause.
Someone may wish to object to the new definition that it is an ad hoc
modification for the sake solely of accommodating deontic logic within
a theory of conditions. My rejoinder will consist of two parts:
First, I do not think the objection a fair one. Quite apart from con-
siderations relating to deontic logic and normative concepts, the suggested
modification seems to me reasonable. But it challenges the serious
problem of how to deal with relations of conditionship between non-
contingent, i.e. necessary or impossible, propositions. Since this problem
is not relevant to deontic logic, we need not discuss it at length here. Let
it only be said that I think its solution has to be given in terms of higher
order (iterated) modalities. A necessary condition of a necessary propo-
sition, I submit, is not a necessary condition of the truth of that propo-
sition, but of its necessity; and a necessary condition of an impossible
proposition is not a necessary condition of the falsehood of that propo-
sition, but of its impossibility.
Secondly, instead of modifying the definition of a necessary condition
we could correspondingly modify the definition of the O-operator by
adding to it the requirement that the states 'p' and 'f' should be contin-
gent. This would seem entirely unobjectionable and would lead to essen-
tially the same results as far as deontic logic is concerned. But, as indi-
cated, I should favour the more daring modification proposed above.
Let it be observed in passing that 'Op~ O(p v q)' is not a theorem of
the deontic logic we are building. This is the famous Ross's Paradox
formula. The fact that it is not a theorem does not mean, however, that
the notorious troubles caused by this paradox, and its variations, have
been completely overcome. For, under a conditional clause to the effect

that the disjunction 'p v q' is not necessary, i.e. the conjunction
'", p &'" q' is possible, the formula becomes derivable. It is easy to find
examples of propositions which satisfy this clause, - e.g. the stock
example 'if it ought to be the case that this letter is mailed, then it also
ought to be the case that this letter is mailed or burnt' will obviously satisfy
it.
The fact that Ross's Paradox formula is not a theorem reflects a re-
striction on the validity of the distribution principle for the O-operator.
It does not follow logically from the fact that the conjunction of two
states is contingent that each one of the states individually is contingent;
one of them may also be necessary. The implication 'O(p &q)~ Op&Oq'
therefore holds only subject to the conditional clause 'M '" p&M '" q'.
But if both of two contingent propositions are strictly implied by a third
contingent proposition ('1'), then their conjunction will necessarily be
contingent. The implication 'Op&Oq~ O(p&q)' therefore holds un-
conditionally.
IV
I now proceed to the sufficient condition aspect of deontic logic.

The definition of the notion of a sufficient condition which corresponds
to the definition which was first suggested for the notion of a necessary
condition is as follows:
'Seep, q)' = df 'N(p ~ q)'.
A sense of the phrase 'it may be the case that p' is then defined as
follows:
'Pp' = df 'Seep, I)'.
According to this suggestion, that something may be the case, or is
'permitted' in a loose and wide sense of this term, means that the thing
in question is a sufficient condition or guarantee of a certain state 'J'.
What this 'J' is will for the time being be left open.
The symbol 'P' will henceforth be used only for the above sense of
'may', and not for the sense which is also expressed by ',.., 0 ",'.
From PL and M and the suggested definitions of 'Se' and 'P' we easily
prove the distribution law 'pep v q)+-+Pp&Pq' or that it may be the
case that p or q, if and only if, it may be the case that p but also may be
the case that q.
On the basis of considerations of symmetry alone, but also for other

reasons, it may be thought that, if the relation of necessary conditionship
is restricted by definition to contingent propositions only, the relation of
sufficient conditionship should be similarly restricted. We achieve this
restriction through the following modification of the definition:
'Seep, q)' = dl 'N(p ~ q) & Mp & M ,.., q'.
Both under the unmodified and the modified versions of the definitions
of 'Ne' and 'Se' these equivalences hold true: 'Se(p, q)+-+Ne(q,p)' and
'Seep, q)-Se(,.., q, ,.., p)' and 'Ne(p, q)-Ne(,.., q, ,.., p)'.
Accepting the modified definition of 'Se', we have to note a restriction
on the validity of the P-distribution principle, analogous to the restriction
we noted in Section III for the O-distribution principle. It does not follow
logically from the fact that the disjunction of two states is contingent that
each one of the states individually is contingent; one of them may be
impossible. The implication 'pep v q)~Pp&Pq' therefore requires a
conditional clause 'Mp&Mq'. On the other hand, if both of two contin-
gent propositions strictly imply a third contingent proposition ('I'), then
their disjunction will necessarily be contingent. The implication 'Pp &Pq
~P(p v q)', therefore holds unconditionally.
Assume that 'p' is something which, in the strong sense, may be and
that 'q' is something which ought to be. Then we have, according to our
definitions, 'N(p~I)&N(I~q)'. This entails 'N(p .... q)' which entails
'p~q'. Thus if something which, in the strong sense, may be the case
actually is the case, then everything which c!lght to be the case is the case,
too. Of that which is in the strong sense permitted one can, in other
words, avail oneself only on condition that none of one's obligations is
thereby violated. This is a refiexion, in deontic terms, of the general
principle for conditions which says that a sufficient condition of some-
thing can be realized only provided that all the necessary requirements
for the occurrence of the thing in question are satisfied as well. To say
of something that it is in the strong sense permitted without including
in the description mention of all dutybound things is therefore an elliptic
mode of speech. Similarly, to say that something is a sufficient condition
of something else without mention of the necessary conditions is elliptic,
too. This elliptic mode of expression is, however, quite commonly used.
Accepting the modified definitions of the notions of necessary and of

sufficient condition, it is easily shown that 'Pp -+ '" 0 '" p' is a theorem.
If something may be, then it is not the case that its contradictory ought
to be. Now, as we have already noted, there is a weak sense of 'may'
which means simply that whatever is such that its contradictory is not
required (for a certain thing) 'may' be. This is the notion of permittedness
which figures in the traditional calculi of deontic logic. The fact that it is
entailed by our new P-notion justifies us in calling the latter the strong
and the former the weak notion of 'may' (or of permittedness).
'Op -+ ~ 0 '" p' is another theorem. If something ought to be, then it
may be - in the weak sense of 'may'. But may it also be in the strong
sense? The answer is negative.
'", 0 '" p -+ Pp' is not a theorem. This observation is related to a
problem which has been very much discussed in the theory of juris-
prudence. The problem is whether anything which is not prohibited
(' '" 0 '" p') is ipso facto permitted. The principles of legal philosophy
couched in the Latin words nullum crimen sine lege and nulla poena sine
lege seem to favour an affirmative answer. Accepting this answer for the
whole of a given legal order would mean that this order is closed, has no
'gaps' or 'lacunae' in it. Some legal philosophers have accepted this and
some have even thought that a legal order is of necessity closed. This
appears to be the position, e.g., of Hans Kelsen. But most legal theorists
seem to think that the closed or open nature of a legal order is a matter
of contingent fact. A difficulty for those, who follow this line of thought,
has been to show why it is that - speaking in the terms of deontic logic -
'Pp' does not follow logically from ,~ 0 ~ p'.
Consider a set of contingent propositions S and a single contingent
proposition 'p' which is not a member of S nor a truth-function of mem-
bers of S. Assume that 'p' has at least one necessary condition which is a
member, or truth-function of members, of S. Assume further that the
conjunction of all necessary conditions of 'p' which are members, or
truth-functions of members, of S is a sufficient condition of 'p'. If these
assumptions are satisfied, I shall call the proposition (state of affairs)
'p' determined in S.
If 'p' is determined in S then, if 'p' is true (obtains), some one of its
sufficient conditions obtains too. The idea that, for a given state 'p', there
exists some set S in which 'p' is determined captures, I think, an impor-
tant aspect of the notion of determinism. Similarly, the idea that, for all
contingent states, there is some such set captures an aspect of the idea
of Universal Determinism. s
Consider next a normative code or order or system which stipulates
that certain contingent states, 'p', 'q', ... , are necessary conditions of a
contingent state 'I' which is not itself one of those contingent states, nor
a truth-functional compound of some of them. Assume that the conjunc-
tion of all those necessary conditions of 'f' is a sufficient condition of '1'.
'I' is determined, in other words, in the set of obligations stipulated by
the code (in terms of 'I'). This assumption means that, if everything that
according to this code ought to be the case actually is the case, then
whatever else there is (happens to be) may be the case. Or, speaking in
juristic terms: if all obligations are satisfied, then whatever else is done
(or omitted) is also allowed. Then we shall say that the normative code
or order or system is closed with regard to the obligations defined in the
terms of 'J'. This idea of the closedness of a legal (or other normative)
order is thus a special case of a more general idea of determinism.
An example will be given to illustrate the above. Assume that 'I' is
determined in some S and that the conjunction of all its necessary con-
ditions in S is equivalent with 'p&q'. Then we have 'O(p&q)' and
'P(p&q)'. Consider now a state Or' which is, as I shall say, co-contingent
with 'p&q'. By this I mean that 'M(p&q&r)' and 'M(p&q&- r)' are
both ofthem true. Then we can derive 'P(p&q&r)' and 'P(p&q&- r)'.
Of course, neither or' nor its contradictory' - r' is by itself a sufficient
condition of 'J'. cPr' and 'P"" r' are not derivable. But we can introduce
here a new piece of terminology and say that, if a state of affairs is co-
contingent with the conjunction of all the necessary conditions of a
determined state 'I', then this state and also its contradictory state are,
in the frame or setting of the necessary conditions, sufficient conditions
of 'J' and in this sense also strongly permitted. (Cf. what was said in
Section IV on the elliptic use of 'permitted' and 'sufficient condition'.)
Since, moreover, neither the state in question nor its contradictory is
itself a necessary condition of '1' - this would conflict with the assumption
of co-contingency - both states are in the weak sense permitted. '"" 0 "" r'
and '- Or' are derivable.
VI
A notion which has caused students of deontic logic considerable trouble

is that of commitment. It is related to the notion of a conditional obli-
gation (norm).
A 'formalization' of commitment which has been suggested is
'0 (p - q)'. The formula can be read 'it ought to be the case that, if it is
the case that p then it is also the case that q'. If it is in the power of an
agent to produce the two states ('at will'), then this agent will, by pro-
ducing the first state, commit himself to producing the second state as
well. This means: unless he now produces the second state, he will have
done something forbidden.
The suggested 'formalization' is open to the objection that, if the first
state is, in itself, forbidden and the second is not in itself, necessary, then
doing the first commits one to doing the second. From '0 ~ p &M ,..., q'
follows '0 (p - q)'. This is, in fact, simply a version of Ross's Paradox.
Another suggestion as to how to 'formalize' commitment, is 'p - Oq'.
This evades the paradox that doing the forbidden 'commits' one to doing
anything whatever (which is not in itself necessary). But then we run into
the new 'paradox' that relative to that which is not (whatever it be) any
other thing o\lght to be.
When deontic logic is placed in the setting of a theory of conditions,
these difficulties can be overcome. Consider first the following point
about necessary conditions:
A necessary condition can have disjunctive form. That 'p or q' be the
case can be a necessary condition of 'r', say. Now assume that 'p' is not
the. case. Then, relative to this, 'q' so to speak 'becomes' necessary, if
'r' is to occur at all. How shall this be expressed? 'Nc(p v q, r)' by itself
cannot express it. That 'q' is a necessary condition of'r', if 'p' is not
there, entails (presupposes, requires) that 'p v q' is a necessary condition
of or', but is a stronger statement. '~p - Nc(q, r)' will not do either. For
it says that, when 'p' is not, then any state is a necessary condition of'r'
- and this is certainly not intended. Nor does this expression entail
'Nc(p v q, r)'.
Consider, however, 'Nc(Nc(q, r), ,.,., p)' or, which means the same,
'Sc( ~ p, Nc(q, r)'. The last formula says that the absence of the state
'p' is a sufficient condition for the state 'q' to be a necessary condition of
the state 'r'. If we make the assumption that it is possible for both the
two states, 'p' and 'q', to be absent, i.e. if we assume that their disjunction
is not in itself a necessity, then the above second order expression entails
that the disjunction of the state 'p' and the state 'q' is a necessary con-
dition of the state 'r' - though not viee versa. We have, in other words,
'Se( - p, Ne(q, r»)&- N(p v q)~Ne(p v q, r)'.
Now substitute 'J' for Or'. Then we get from the last formula
'See - p, Oq)&- N(p v q)~ O(p v q)'. The antecedent, I suggest, is an
expression for commitment (relative, conditional obligation). I shall
define a new deontic operator Q as follows:
'Q(qjp)' = df'Se(p, Oq) & - N(p ~ q)'.
'Q(qJp)' can be read 'it ought to be the case that q, given that p'.
Assuming that it is in the agent's power to produce 'p', then by producing
this he 'becomes' obligated to produce also 'q', unless 'q' is something
which is of necessity there as soon as 'p' is there. (For in this last case,
it 'makes no sense' to speak of an 'obligation'.)
'Q(qjp)' entails 'O(p~q)' but not viee versa. 'O-p' does not entail
'Q(qjp)'. Nor does 'Oq' do this.
Of particular interest is the case when '0 - p' actually obtains with
'Q(qjp)'. Then the second expression tells us what the agent ought to do,
when he has done something he was, in fact, forbidden to do. This sort
of commitment or conditional obligation is of the kind for which Pro-
fessor Chisholm has coined the name Contrary-to-Duty Imperative. 6
The above ideas seem to me to lead to a satisfactory account of the
notion of commitment, and of conditional norms generally. Some
problems of old standing in deontic logic now acquire a natural solution. 7
VII
The incorporation of deontic logic into a modal logic of conditions also

opens new prospects for dealing with problems relating to the iteration
of deontic operators. In 'traditional' deontic logic these problems are
notoriously obscure.
That deontic operators can become iterated is' no more of a 'mystery'
than that conditions can be of higher order. The concept of commitment,
and of a conditional norm generally, is - as we have just seen - a second
order condition concept. In it is involved the idea that the fact that
something is a necessary condition of something else is itself a necessary
condition of something.
If we let these things of which something is a necessary condition be
identical with the state of affairs 'J' which figures in our definitions of the
deontic operators, we obtain higher order deontic expressions such as
'OOp', 'OPp', 'POp', 'PPp'. Thus, for example, 'OOp' when explicated
in terms of conditions means 'Nc(Nc{p, I), I)'. When further explicated
in modal terms we get from this the expression 'N(I --.. N{I --.. p) &MI &M
"'p)&MI&M", (N(I--"p) &MI&M "'p)'. If we ignore, or take for
granted the contingency-conditions attached to the antecedents and
consequents of the two strict implications involved, we can simplify the
formula to 'N(I--..N{I-+p)'.
In the light of this interpretation of higher order deontic expressions
one can profitably examine some formulae, whose status as logical truths
about the deontic modalities has been defended by some and disputed
by others. I shall consider one example. This is a formula which I shall
label Prior's Formula. 8
It is clear that 'Op --.. p' cannot be a truth of logic. (Cf. Section I.) But
what about 'O{Op-+p)'? In words this formula says that it ought to be
the case that that which ought to be the case actually also is the case. This
sounds reasonable enough. Let us see what it amounts to when the
formula is translated into the terminology of conditions and modalities.
As a statement of conditionship the formula means 'Nc({Nc(p, I)
--.. p), I)'. If we write it out in modal terms we get 'N(I --.. (N(I --.. p) &MI
&M,.., p--..p) &MI&M(N{I--..p)&MI&M ""p&"" p)'.
Let us first ignore the contingency-clauses and 'pull out' the strict im-
plication part of the formula. Then we get 'N(I --.. (N(I --.. p) --.. p)'. This
is easily shown to be a theorem of modal logic (M). Therefore the condi-
tional statement which says that, if the contingency-clauses are satisfied,
then Prior's formula holds good, is a theorem of modal logic, too.
We now tum attention to the contingency-clauses. The third member
of the main conjunction in the modal expansion of Prior's formula can
be abbreviated to 'M ( Op &,.., p)'. It says that it is possible that the state
of affairs 'p' ought to be but nevertheless is not.
But is this clause needed at all? Is it not trivially satisfied by virtue of
the fact that 'Op --.. p' is not a theorem? The answer to the last question
is negative. The mere fact that 'Op-+p' and therewith also 'N(Op-+p)'
is not a theorem does not make the negation of 'N(Op-+p)' which is
equivalent to 'M(Op&- p)' a truth of modal logic. Whether the clause
upon which Prior's formula is conditional is satisfied or not is therefore
a matter of contingent truth and not of logical necessity.
What the 'if-clause' which is needed to warrant the truth of Prior's
formula does, interestingly, is to draw attention to the fact that, although
it is inherent in the notion of an obligation that the obligatory is some-
thing in itself contingent and in this sense neglectable, it may yet as
obligatory be impossible to neglect. cOp -+ M - p' is a theorem of deontic
(modal) logic. But 'Op&M(Op&- p)' can be true or false, and so can
'Op&N(Op-+p)'. When 'Op&M(Op&-p)' is true ofa state 'p', J shall
say that the fact that 'p' ought to be is a neglectable obligation. When
again cOp & N (Op -+ p)' is true of a state I shall say that the fact that this
state ought to be is a non-neglectable obligation.
I conclude this section with the conjecture and sugcstion that the
distinction which we have discovered between two types of obligation
can be interestingly related to things familiar from traditional ethical
theory. (Over-riding obligations, primajacie obligations, etc.)
VIII
In our definitions of the deontic notions we have so far employed an

unspecified propositional constant 'J'. Something will now have to be
said about its possible content.
One suggestion could be that no specification at all of the content of
'J' is needed for a definition of 'ought'. According to this view, to say
that something ought to be or oupt to be done is to say that the being
or doing of this thin. is a necessary condition of a c:ertain other thing
which is taken for granted or presupposed in the context. An 'ought'-
statement is typically an elliptic statement of a necessary requirement.
The same holds, mutatis mutandis, for the two types of 'may' statement.
(This elliptic use, omitting reference to a specified 'J', must not be con-
fused with the elliptic use mentioned in Sections IV and V above, when
reference to a frame of necessary conditions is suppressed but tacitly
presupposed. )
This suggestion seems to me, on the whole, acceptable. If we accept it,
then we are always, when confronted with an 'ought', entitled to raise

the question 'Why?', i.e. to ask for the thing for which this or that is
alleged to be a necessary requirement. There may exist a tendency,
particularly in so-called moral contexts, to forget about the elliptic
character of the 'ought', and to accord to it an absoluteness which in-
trinsically it cannot have. It is an important aspect of the task of the
moral critic to challenge accepted 'oughts' by raising for them the
question 'Why?' or 'What for?'
Even if we accept that the 'J' in our definition of 'ought' can be any
state whatever, which is capable of having necessary conditions, there
are some types of case which are worthy of special attention. The distinc-
tions between the types have partly to do with the content of 'J', and
partly also with the kind of necessity involved in the contemplated
conditionship-relation.
One important type of case is when 'p' is thought to be a logically
(conceptually) necessary requirement of '1'. For example: A good
fountain-pen, we say, ought not to leak. Or: In order to be legally valid
a marriage ought to be concluded in the presence of two witnesses. A
leaking fountain-pen does not count or qualify as a good one - and a
marriage at which two witnesses have not been officially recorded does
not count or qualify as valid. But the standards of legal validity may be
different under different legal systems; and also standards of goodness
in fountain-pens can vary depending upon preferences.
Another characteristic type of case is when the state 'J' is the aim or
end of some agent for the attainment of which the being or doing of 'p'
is thought to be causally necessary. The thing which ought to be or be
done is then required as a necessary means to the given end. An example
would be when I say: 'in order to get there in time I ought to hurry'.
StipUlations or rules conforming to the first pattern of 'ought' are or
resemble definitions. Rules or norms in accordance with the second
pattern of 'ought' can be called technical norms or alternatively practical
necessities.
The 'may', both in its weak and strong sense, can also be of one of the two
types which we just described. If I say that a good fountain-pen need not
be of gold but may be of some material other than gold, I mean that this
particular requirement concerning the material of which it should be
made is not built into the (my) concept of a good fountain-pen. This is
the weak 'may'. Similarly in the case when I say that in order to get to a
destination in time I need not take a tram, but may also use some other
means of transportation. Assume, however, I said that in order to get
there in time, I may take a bus. This would normally mean that, if I go
there by bus I shall arrive in time, but that there perhaps are other means
beside this to secure the same end of my action. I may, e.g., also take
the subway, I can choose between it and the bus. This is the strong 'may'.
The problem of the content of 'j' is of particular interest to philosophy,
when the 'ought' (and the two 'may') is a legal or a moNI 'ought' ('may').
I would submit for consideration that for an important type of lepl
'ought' - the 'ought of lepl obligation' as I shall call it - 'j' is a state of
affairs which can be characterized as immunity to pIIfIislunmt (a punitive
reaction on the part of a legal machinery). The actions which it is our
legal duty or obligation to do are those which are required of us if we
are to be immune to lopl punishment. They are necessary to ensure
immunity, but whether they are also sufficient depends upon whether the
legal order, or part of legal order, under consideration is closed or open.
(Cf. above Section Y.)
But what is immunity? This is itself a problematic notion and can be
understood in several ways. One could suggest this: Immunity means
that, unless 'p' is (done), punishment will follow. But this would hardly
be an interesting notion of immunity. For 'p' may be neglected and yet
no punishment follow, either because the criminal is not caught or,
although caught, acquitted because not proved guilty. To be immune to
punishment for neglect of 'p' means rather that one cannot be legally
punished for having neglcctecl 'p'. And the contradictory of immunity,
which can also be called liability to punishmeDt, means that one can be
thus punished. But what do 'can' and 'cannot' mean here? They are
modal notions and on that gt'ound alone related to 'ought' and'may'.
In the analysis of the notions of immunity and liability to punishment,
deontic notions may thus crop up anew. This would not entail circularity
or show that immunity and liability cannot be interestingly used for
defining the notions of legal obligation and permission. But it would
show that in the structure of a legal order other types of 'ought' and
'may' than (legal) obligations and permissions· are involved - and it
would chal1enge further analysis of how the various types of norm which
build up this order are related and intertwined.
IX
Condition-statements, for example to the effect that something is a

necessary or is a sufficient condition of something else, are genuine state-
ments. By this I mean that it makes sense to attribute truth-value to them,
to speak of them as being true or false. It does not follow that they are
'naturalistic', if by that is meant that the grounds on which they are
pronounced true or false are experiential (empirical). When the condition-
statement is about the necessary or sufficient means to some end, its
truth-grounds are - normally at least - experiential and the statement
'naturalistic'. When, on the other hand, the statement is about the logically
necessary requirements for falling under a concept (qualifying as a such-
and-such), the truth-grounds are not experiential - and there need not
exist any truth-grounds for the statement at all.
'Ought' and 'may' have an important connexion with norms. Norms
are not ordinarily called 'statements' or 'propositions'. Whether truth-
value can be attributed to norms is a matter of controversy. Many
philosophers and logicians have thought that norms are essentially void
of truth-value, 'outside the realm of truth and falsehood', belong to
'practical' as distinct from 'theoretical' discourse. And some of these
philosophers have therefore thought that norms are essentially a-logical,
that there can be no such thing as a 'logic of norms' or deontic
logic.
It is important that the 'theoretical' character of condition-statements
should be reconciled with the 'a-theoretical' aspect of norms - the 'ought'
and 'may' which express necessary or sufficient conditions with the 'ought'
and 'may' of genuine norms. Such reconciliation is, I think, fully possible
- and to effect it is a philosophically relevant task. A satisfactory accom-
plishment of it would remove much confusion which has of old prevailed
in legal and moral philosophy.
Why is it that to say that something ought to (or may) be the case or
be done so often has the appearance of not being a genuine statement,
to which a truth-value can be significantly attached?
There are several reasons why this is so. - One is, I think, to be sought
in the elliptic character of many 'ought' - and 'may'-statements. If some-
one says 'this ought to be' and another 'that ought to be' and the 'this'
and the 'that' exclude each other and there is a dispute, we cannot begin
trying to settle it until we have first stated/or what the 'this' and the 'that'
are thought to be required. If they are required for different things, there
can be no dispute between the two 'oughts' (but perhaps another dispute
about which ofthese two things ought to be pursued). If they are required
for the same thing (the same '1'), one of the disputants may be right and
the other wrong or both may be wrong. And now 'right' and 'wrong' is
a matter of truth and falsehood.
Even when the elliptic character of a given 'ought' (or 'may') is recog-
nized in principle it may be difficult or even impossible to specify in
practice the '1' relative to which something ought to or may be. Perhaps
we only have some dim conception of its nature. (We had been taught to
think the matter was clear without questioning. This is what all too often
happens in an authoritarian type of society.) Then we may not come to
a 'grip' with the question of truth in connexion with the 'ought' ('may')
and the norm takes on an 'alogical' appearance.
A characteristic use of 'ought' (not so much it would seem of 'may')
is to evince an evaluation. Thus in saying "a good fountain-pen ought
not to leak" or simply "a fountain-pen ought not to leak", I may be
evincing a standard of goodness in fountain-pens. "A leaking fountain-
pen, according to my conception ofthe matter, simply is not a good one."
To adopt this standard is to make the property of not leaking a necessary
requirement of goodness in fountain-pens.
Another characteristic use of 'ought' (and to some extent also of the
strong 'may') is for enforcing patterns of behaviour (conduct). This is,
in a paramount sense, the normative use of the words 'ought' and 'may'.
Imperatives (sentences in the imperative mood) are also used for the
same purpose. It would be a confusion to say that 'ought'-statements are
imperatives. Imperatives are not statements and imperatives, as has so
often been pointed out, are not true or false. But 'ought'-sentences are
commonly and characteristically used as imperatives, viz. used for the
purpose of urging (making) people to behave in a certain way.
It is futile to dispute whether 'open the window' and 'you ought to
open the window' mean (as it were 'intrinsically mean') the same or not.
Sometimes someone addresses another person with the words "you ought
to open the window", when he could just as well have said "open the
window". But perhaps he will, when he says the former, more often than
not have 'at the back of his mind' an idea of the thing for which com-
pliance with the order is a necessary requirement, - e.g. that ventilation

of the room is badly needed. And one can be pedantic and say that 'ought'
is properly used only when there is an answer to the question "Why?",
whereas use of the imperative can also be appropriately made when this
question is out of place.
The use of 'ought' for giving commands and orders and the use of
'may' for giving permissions is a species of performative use of language.
Language, when used performatively, can rightly be said to be "outside
the categories of truth and falsehood". But the same sentence which has
a characteristic performative use can also have a characteristic descriptive
use. And the two types of use may fuse.
Finally, some peculiarities on linguistic usage will be noted.

When we say '''p' and 'q' and ... is a necessary condition of _" we
intimate that 'p' and 'q' and ... are all of them, individually, necessary
conditions of_.
When we say '''p' or 'q' or ... is a necessary condition of _", we
intimate that 'p' and 'q' and ... are none of them individually, nor some-
but-not-all of them disjunctively, a necessary condition of_.
When we say "'p' and 'q' and ... is a sufficient condition of _", we
intimate that 'p' and 'q' and ... are none of them individually, nor some-
but-not-all of them conjunctively, a sufficient condition of_.
When we say "'p' or 'q' or ... is a sufficient condition of _", we
intimate that 'p' and 'q' and ... are all of them, individually, sufficient
conditions of_.
The second and the third case, i.e. 'or' in connexion with necessary and
'and' with sufficient conditions, bear on anomalies of the type of Ross's
Paradox.
The 'and' and 'or' in these locutions are not identical with the truth-
functional notions of conjunction and disjunction.
Compare 'ought to', 'should', 'must', 'has to'. Is not 'must' stronger
than 'ought to', more expressive of a necessary requirement (condition)
than 'ought to'? So it is, but I do not think one can make hard and fast
distinctions here. Even necessity can possess degrees of rigour or laxity.
Sometimes higher order modal concepts can be used for expressing this.
That which is necessarily necessary may in an interesting sense be said

to be 'more necessary' than that which is only contingently necessary.
If there exists any way at all of capturing the distinction between 'ought
to' and 'must' by means of logic, this would have to happen within a
theory of conditionship-relations of different orders. (Cf. what was said
in Section VII about neglectable and non-neglectable duties.)
Academy of Finland
NOTES
• An earlier version of this paper appeared in Critica 2 (1968) 3-25. The author wishes
to acknowledge his indebtedness to Professor Carlos Alchourr6n (Buenos Aires) for
criticisms and suggestions which have been useful in revising the essay.
1 I shall not explain here my use of symbols nor the conventions adopted about
brackets. I shall assume that the symbolism will be either familiar to the reader or else
self-explanatory .
2 I shall use the words 'obligatory', 'obligatoriness', 'permitted', and 'permittedness'
with a loose and wide meaning which is roughly equivalent with ordinary uses of
'ought' and 'may'. There is also a stricter use of the words mentioned and of the terms
'obligation' and 'permission', which belongs typically in legal and moral contexts.
3 'A Reduction of Deontic Logic to Alethic Modal Logic', Mind 67 (1958) 100-103.
4 Another treatment of the elements of a logic of conditions is found in my book A
Treatise on Induction and Probability, Routledge and Kegan Paul, London, 1951. That
treatment is framed, not in modal propositional logic, but in the monadic lower
functional calculus (theory of properties and quantifiers).
5 Cf. A Treatise on Induction and Probability, Chapter Ill, Section 2.
8 R. M. Chisholm, 'Contrary-to-duty Imperatives and Deontic Logic', Analysis 24
(1963) 33-36.
7 Cf. also my paper 'A New System of Deontic Logic' (in this volume, pp. 105-120.
It is readily shown that, subject to the appropriate contingency conditions, the dyadic
expressions 'Q(f)' satisfy the Axioms BI-B3 given in the original version of the paper.
The undesirable consequence which led me to abandon Bl in favour of Bl' does not
arise, if the notion of a conditional obligation is explicated, as here, in the terms of
conditionship relations between contingent propositions. The problem of predica-
ments or conflicting duties will now have to be treated differently, viz. along the lines
suggested in my paper 'Deontic Logic' (Mind 60 (1951) 1-15). If a man, like Jephtha,
commits himself to do something which is forbidden, then the committing action was
itself forbidden. 'Q(q/p)' in conjunction with '0,.., q' entails '0 ,.., p'.
8 It was introduced into the literature by A. N. Prior in his Formal Logic, Clarendon
Press, Oxford, 1955.
INDEX OF NAMES
Alchourron, C. xiv, 177 Foot, P. 103

Anderson, A. R. 14, 19-21, 31, 34, 35, van Fraassen, B. C. xii, xiv
54, 82, 109, 129, 134, 145, 147, Frege, G. 48, 50
153,157,161
Aquinas, T. 12, 34 Geach, P. 28,47,115
Aqvist, L. xiii, 34, 35, 132, 147, 153, Gill, J. H. 103
156, 157 Gombay, A. 32,35
Aristotle 44,48 Greenspan, P. xiii
Ayer, A. J. 57,58 Grelling, K. 7, 9, 10, 15, 32-34, 83
Grice, H. P. 32, 35
Bar-Hillel, Y. 31,35
Bentham, J. 1 Hagerstrom, A. 57
Beth, E. W. 49 Hanson, W. H. 16, 32
Bochenski, I. M. 48 Hansson, B. xii, 8, 21, 28, 31, 32, 34,
Bobano,B.l,48-50 35, 103, 121-147
Broad, C. D. 31,34 Hare, R. M. 58
Bulygin, E. xiv Hedenius, I. 57
Henkin, L. 152
Chisholm, R. M. xii, xiii, 24-29, 31, 35, al-Hibri, A. xii
105-lO9, 115, 119, 120, 132, 147, Hilpinen, R. v, xii, xv, 1-35, 157, 158
169,177 Hintikka, J. v. vi, xii, xv, 14, 16-18,
Carnap, R. 33 32,34,35,59-104,157,158
Castaiieda, H.-N. xv, 31, 34, 137, 147 Hockney, D. J. 31,32,34
CheUas, B. F. xi, xii, xiv, 157, 158 Hoepelman, J. xiii
Church, A. 48 Hofstadter, A. 6,32,40,47,49
Commons, J. R. 43 Hohfeld, W. N. 32,34,35,43
Cook, W. W. 32 Hospers, 1. 58
Cresswell, M. 32,34 Hughes, G. E. 32, 34, 83
Hume, D. 58, 99
Davidson, D. 60,61
Davis, J. W. 31,32,34, 157 1ennings, R. E. xiv
Dawson, E. E. 32, 34, 35 191rgensen, J. 8, 32
Dopp,J. 32
Kamp, H. xiv, xv
Feys, R. 19,32,34 Kanger, H. 32,35
Fisher, M. 32, 34 Kanger, S. v, xii, xv, 16, 19-21, 32,34,
Fitch, F. B. 34 35,36-58,128,129,147
F;llesdal, D. v, xii, xv, 1-35 Kant, I. 49,50,73,74,78,86,87,163
INDEX OF NAMES 179
Kelsen, H. 166 Quine, W. V. O. 1,33,34,50

Klibansky, R. 33,157
Knuuttila, S. xi Rand, R. 6, 33
Kripke, S. A. 16,18,32,155,156 Reach, K. 7,33
Rescher, N. 31, 33, 35, 60, 121, 133-
Lambert, K. 157 136, 138, 140, 141, 144-147
Lemmon, E. J. xi Richards,1. A. 57
Levi, I. 21,32, 35 Robinson, J. 33,35
Lewis,D. xii,xiii,151,157 Ross, A. 8, 21, 22, 33, 131, 132, 163,
Loewer, B. xv 164,168,176
Ross, W. D. 43,58,92, 103
McKinsey, J. C. C. 6, 32, 40, 47, 49 Russell, B. 57,58
Makinson, D. xiv, 152
Mally, E. 1-7,9,11,12,14,15,32,34, Schilpp, P. A. 33,58
39,82,109 Schotch, P. K. xiv
Mates, B. 48 Scott, D. xi, lSI, 152, 157,158
Meinong, A. 1 Searle, J. R. 93-95, 97-99, 103, 104
Melden, A. I. 33 Segerberg, K. v, xii, xiii, 148-158
Menger, K. 6,32 Sellars, W. 58
Montague, R. 16,33,151,157 Stenius, E. 8,14,33,119,120
Moore, G. E. 43,58 Stevenson, C. 58
Moore, O. K. 54 Svenonius, L. 119
de Morgan, A. 69 Swartz, R. 32
Mott, P. xiii
Tarski, A. 46,48,49
Ogden, C. K. 57 Thomason, R. H. xiii
Perry, R. B. 58 Wilson, W. K. 31,32,34

Powers, L. 140, 147 von Wright, G. H. v, xii, xiv, xv, 1,8-15,
Prior, A. N. 7, IS, 20, 23, 24, 33, 35, 19,20,22-24,26-29,31,33-35,39,
47,54,65-67,79,81-83,109,170, 40,49, 83, 105-120, 121, 123, 124,
171,177 127-129, 132-134, 137-141, 144,
145,147,159-177
INDEX OF SUBJECTS
Act - deontic 77-87

generic 9-10 - logical 48-49,68,77-79
individual xv, 10, 14, 59-{;1, 65-67 Contingency 163-165
performance-value of 13-14 - conditions 170-171
predicate 10,13,51,58 Correctness (of imperatives) 37, 45-
Alternative, see Deontic alternative 47,51
Alternativeness relation 17-19, 70-72,
75-77,151-155 Deontic
Analyticity 37, 48-50 - alternativ~ 17-19, 70-77, 102-103,
Axioms for deontic logic 151-155
- Grelling's 7 - consequence 77-87
- Mally's 2 - distance 150
- Rescher's dyadic system 133-135 - implication 12,34,78-79
- Segerberg's, for commitment 152- - proposition 46
156 - sentence 1
- standard system xi, 13, 127-128 situation 156
- von Wright's 8-9, 127-128 see also Situation
- von Wright's dyadic system 27, 110, space 116-117
116-117, 137 Deontic logic
- and action theory xv
Blameworthiness 42-44 and alethic modal logic 15-19,
159-160
Circumstances 26-27,133,139-142 axioms for, see Axioms for deontic
Claim 34-35,130 logic
Commitment xii, xiii, 6, 12, 23-24, descriptive interpretation of formu-
87-101,148-158,168-169 lae (deontic sentences) 8, 122-123
absolute (strong) 99-101, 157-158 - dyadic systems of xii, 26-31, 105-
different senses of 87-101 120,121,132-147
laws of 11-12 and logic of imperatives 1, 45-47,
- logic of 148-158 49,51,55
paradoxes of 23-27, 81, 88-90, - minimal xiv
168-169 - quantifiers in 47,53, 61-{;7, 101-103
prima facie (weak) 99-101,157-158 - reduction to alethic modal logic 19-
Completeness 18, 77 21,53-54,128-129,161
Conditions semantics of xv, 15-19,53-54,67-
logic of xii, 160-161, 177 77, 123-127, 143-146, 150-157
see also Theory of conditions standard system of xi, 13, 122
necessary 162-164, 176-177 Deontic operators
sufficient 160, 164-165, 176 - iteration of 10, 14-15, 73, 169-171
theory of, see Theory of conditions - relative to persons 34-35, 60, 148
Consequence Deontically perfect world 17-19, 26,
30-31,71-75,102-103
Deontik 1-2,39 alethic xi, 8-9, 19-21, 83-87, 159-

Determinism 166-167 160
Duty deontic xi, 8, 19-21, 159-160
absolute (actual) 91-99 relation between deontic and alethic
conflict of duties 118-119 xi, 19-21, 53-54, 83-87, 128-129,
see also Predicament 158-160, 162-177
prima facie 91-99 - relative to persons 34-35,60
temporal xiii
Emotive theory 55,57-58 Model 17, 45-46, 151-152, 155-156
Ethical theory ~t 16,69-73,75-77
- strength of 37 structure 18
- stringency of 37-38 - system 17-18,72,75-77
Evaluation 175 - theory 16
Exposure 43-44 Moral conflicts xiii-xiv
Forbidden 59, 62 Naturalism vs non-naturalism 56-58

see also Prohibition Naturalistic fallacy 56, 58
Freedom 86 Necessity xi, 19, 53-54, 129, 155-156,
159-161
Immunity 42-44 - causal 172
- to punishment 20, 173 Neighborhood 150-153,155-156
Imperative operator 7,37-40 No-right 43
Imperatives 1, 46 Normative logic 53-55
categorical 114-115 see Deontic logic
contrary-ta-duty 25-30, 105-107, Normative systems 1,8,13,16
114-116,119,169 - consistency of 8, 16-19
logic of 1, 6-8, 49 - logic of 1,13
and ought 175-176 see Deontic logic
Implication - open vs closed xiv, 166
deontic 12, 34, 78-79 Norms
logical 5,79 - conflicts of xii-xiv
material 6, 12, 90 - logic of, see Deontic logic
strict 5-6,12,133-134 satisfaction of 49
- semantics of 67-77
Judgement 45-46 see Deontic logic, semantics of
technical 172
Kingdom of ends 73-75,86 theory of, see Deontic logic
Kripke model 155-156
O-constituent 111-112
Liability 42-44 Obligation
- to punishment 20, 173 actual (absolute, categorical; as
Liberty 42-44 oppo~d to prima facie) 91-99, 104
absolute (as opposed to conditional)
May, see Permission xii, 28-29, 31, 111
Modallogic 8,68, 159-161 conditional (hypothetical) 27-29,
see also Modality 109-120,133-147,168-169
Modality (modalities) contrary-to-duty 132-133
182 INDEX OF SUBJECTS
see also Imperatives, contrary-to-duty extensionality 9·-11, 106, 115, 117

neglectable vs non-neglectable 171 fait accompli 7
- prima facie 91-99,104,171 permission 9, 11
relativity to time xiii Privilege 42-43
Ought 1-2,37-39 Prohibition xi, 59, 63-64, 100
and can ('ought implies can') xiv, Promising 29, 93-98, 119-120, 157
40-41,83-87,163 Proposition
- deliberative vs judgmental xv deontic 46
elliptic character of 171-172, 174- - ordinary 46
175 vs practition xv
and is (,derivation of ought from
is') xii, 4-5,93-99, 103-104 Quantifiers in deontic logic 47, 53,
- normative and imperative use of 61-67,101-103
174-176
- and performative use of language 176 Range 44-45,50-51
practical concept of xv Realization 46
Right 1,34-35, 39,41-44, 130
P-constituent 11 - different senses of 41-44
Paradoxes in deontic logic 21-26, Rule of extensionality, see Principle
130-133,163-164 of extensionality
- of commitment (contrary-to-duty
imperatives, derived obligation) xii, Sanction 19-21
23-26, 81, 88, 107-109, 132-133, Satisfiability 16-18,68-73,75-77
168-169 Semantics, see Deontic logic, semantics
of disjunctive permission xiv-xv of
of the Good Samaritan 102 Sentence
- Ross's 21-23, 131-132, 163-164 deontic
Permission (permissibility) xi, 16-17, - imperative 39
59-60,69-70 - normative 39
- conditional 27,133-134 Similarity 149-150, 153, 155
- disjunctive xii, xiv-xv Situation 149-150, 151, 153, 155-156
free choice xiv, 22-23, 160 State ofaffairs 1-2, 14, 17
principle of 9 - generic 105
strong vs weak xii, xv, 160, 165-167 Statement
Possibility xi, 8-9, 159-160, 161 - deontic 39
Possible world 17,30,69,101,123- - ordinary 39
124,129-130,143-144,151 System 45
Power 42-44
Praiseworthiness 42-44 Theory 124-126
Predicament 29,118-119,177 - basisof 125-127
Preference Theory of conditions xii, 160-165, 177
- logic of, and deontic logic 31 - and deontic logic xii, 159-177
- relation 144 Theory-set 125-127
Principle of
consistency xii, xiv Utilitarianism 41
'continuous moral tectitude' 7
deontic contingency 9, 13 Vagueness 51
deontic distribution 9, 11 Validity 18, 48-49, 68, 75-77, 121
Valuation 44-48, 121, 123-124, 143- World

144 - deontically perfect 17-19,26,30-31,
- primary 4, SO-51 71-75,102-103
- secondary 44-45 - ideal 17, 26, 30-31, 129-130, 143,
- possible 17, 30,69, 101, 123-124,
Welfare program 40-41 129-130,143-144,151
Wrong 39

Deontic Logic: Introductory and Systematic Readings

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Deontic Logic: Introductory and Systematic Readings

Uploaded by

Copyright:

Available Formats

DEONTIC LOGIC: INTRODUCTORY AND SYSTEMATIC READINGS

Department of Philosophy, University of Turku, Finlllnd

D. REIDEL PUBLISHING COMPANY

Main entry under title:

(Synthese library; v. 33)

First published in 1971

Sold and distributed in the U.S.A. and Canada

In all other countries, sold and distributed

D. Reidel Publishing Company is a member of the Kluwer Group

First published in 1971 in hardbound edition

All Rights Reserved

and copyright owners as specified on appropriate pages within

PREFACE TO THE SECOND IMPRESSION VII

PREFACE TO THE FIRST EDITION IX

INTRODUCTION TO THE SECOND IMPRESSION xi

DAGFINN FI/>LLESDAL and RISTO HILPINEN / Deontic Logic:

STIG KANGER / New Foundations for Ethical Theory 36

JAAKKO HINTIKKA/Some Main Problems of Deontic Logic 59

GEORG HENRIK VON WRIGHT / A New System of Deontic Logic 105

BENGT HANSSON / An Analysis of Some Deontic Logics 121

KRISTER SEGER BERG / Some Logics of Commitment and

GEORG HENRIK VON WRIGHT / Deontic Logic and the Theory of

INDEX OF NAMES 178

INDEX OF SUBJECTS 180

This second impression of Deontic Logic: Introductory and Systematic

I wish to express my thanks to all contributors to this volume for their

(A com C) :::> «A & B) com C)

(where 'com' stands for the concept of commitment) entailed by

light of Kamp's discussion, the problem of disjunctive permission seems

[20] Loewer, Barry, 'Counterfactuals with Disjunctive Antecedents', Journal of Philosophy

DEONTIC LOGIC: AN INTRODUCTION·

I. THE SUBJECT-MATTER OF DEONTIC LOGIC

II. MALLY'S SYSTEM OF 'DEONTIK'

Most of the contemporary interest in deontic logic has been stimulated

judgment; it lays down the criteria of correct and incorrect judgment.

(AI) (p => Oq) & (q => r) => (p => Or)

(A4') is not well-formed. Mally's interpretation of (A4') can be expressed

Most of these consequences appear fairly plausible from the intuitive

(8) Op&(p::::l q)::::l Oq (from (AI) by (A6»)

(11) Ou ::::l (q ::::l Oq) (from (10»

some sort of strict implication, not material implication. For instance,

III. MALLY'S SUCCESSORS

that is, if we do something we ought not to do, we ought to do it. This

logical consequence, etc., to imperative sentences. Vie'Ns of this kind

IV. VON WRIGHT'S SYSTEM OF DEONTIC LOGIC

The publication of G. H. von Wright's classic paper 'Deontic Logic' [46]

Von Wright's system includes three 'principles' or axioms, which can be

(27) F{PflJ Pf2, ... , Pfm) ,

and by virtue of the principle of deontic distribution, (28) is equivalent to

(37) ...., (0 (p v q) & ...., Pp & ...., Pq)

Von Wright calls these laws "the laws of commitment". According to

(33) can also be expressed in the form

It is interesting to compare this principle with Mally's theorem (9). The

v. THB STANDARD SYSTBM OF DBONTIC LOGIC

Most systems of deontic logic include the system described above as a

If (02) is replaced by (41), we obtain a system equivalent to (0IH03).

pretation of the formalism of propositional logic involves certain diffi-

(44) O(Op ::J p)

VI. SEMANTICAL APPROACHES TO DEONTIC LOGIC

Most of the work described above represents the syntactical or axiomatic

means only that p is compatible with all obligatory states of affairs.19

(Fl) If B = {all> 012, ... , Oln' Pg} holds in M, there is a world

According to (FI), Ml is a world in which all those propositions are true